Copyright TechTarget - All rights reserved ComputerWeekly’s best articles of the day https://cyber.law.harvard.edu/rss/rss.html Techtarget Feed Generator en Thu, 25 Jun 2026 02:37:01 GMT https://www.computerweekly.com editor@computerweekly.com <p>The competitive edge gained from <a href="https://www.techtarget.com/whatis/feature/12-of-the-best-large-language-models">artificial intelligence (AI) model innovation</a> is becoming a temporary rather than long-term advantage for technology suppliers and users. Foundational capabilities are converging, meaning the leaderboards are changing every quarter, according to Gartner distinguished vice-president analyst Arun Chandrasekaran.</p> <p>Speaking during a media briefing on current trends in AI at Gartner’s Data &amp; Analytics Summit in Sydney, Chandrasekaran noted that beyond raw model power, major AI research labs are focused on improving the capability of models in orchestrating workflows and processing multimodal inputs such as speech and images.</p> <p>Highly capable <a href="https://www.techtarget.com/searchcustomerexperience/definition/speech-recognition">speech-to-text</a> models have already emerged, enabling the creation of advanced <a href="https://www.computerweekly.com/news/366641314/How-voice-AI-is-transforming-customer-service">voice agents</a>. Chandrasekaran said it is increasingly likely that when customers call a company, they will be answered by a reasoning, autonomous voice agent.</p> <p>There is also a growing geographical divide in how AI models are being developed and deployed. While AI model innovation in the West remains largely proprietary, there is a broader trend towards <a href="https://www.computerweekly.com/news/366606012/How-open-source-is-shaping-AI-developments">open weight and open source developments</a> in the East. Following a recent tour of six Asia-Pacific countries, Chandrasekaran observed a significant number of open models being used in regional enterprise pilots and production environments.</p> <section class="section main-article-chapter" data-menu-title="Will SaaS survive AI disruption?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Will SaaS survive AI disruption?</h2> <p>Addressing the much-hyped death of software as a service (SaaS) in what has been dubbed “<a href="https://www.techtarget.com/searchitoperations/news/366639662/SaaSpocalypse-Maybe-not-but-SaaS-applications-are-changing">SaaSpocalypse</a>”, Chandrasekaran acknowledged that SaaS is being disrupted by AI, but argued the short-term impact has been exaggerated.</p> <p>SaaS providers retain critical advantages that enterprises rely on, including deep domain knowledge, tight workflow integration and compliance capabilities, he added, noting that for most IT departments, suddenly throwing away a trusted SaaS ecosystem seems far too risky.</p> <p>However, <a href="https://www.techtarget.com/searchcustomerexperience/feature/AI-pricing-is-turning-into-a-software-management-problem">pricing remains the Achilles’ heel of the SaaS sector</a>. Traditional seat-based or user-based pricing models are under threat because AI agents enable fewer people to accomplish the same amount of work. Consequently, suppliers are struggling to balance stock market growth expectations with the shifting operational demands of their customers.</p> </section> <section class="section main-article-chapter" data-menu-title="The changing economics of AI"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The changing economics of AI</h2> <p>The broader economics of AI are also facing intense scrutiny, particularly following reports of companies <a href="https://www.computerweekly.com/news/366644133/Interview-Pegasystems-Don-Schuerman-on-how-to-keep-the-lid-on-skyrocketing-AI-costs">burning through their entire annual AI budgets</a> in a single quarter.</p> <p>Chandrasekaran’s take is that generative AI (GenAI) was adopted by users in their personal lives long before it became a staple of enterprise IT. While AI companies have an incentive to serve the consumer market for brand awareness and to establish a foothold inside enterprises, there is pressure to make business customers pay their way.</p> <p>That’s partly because more businesses are incentivising employees to increase AI consumption – including through agents capable of making sequential or parallel requests, which drives up the total volume of model queries.</p> <p>Since AI providers are constrained by compute capacity, their options for managing this load are limited: they must either reserve capacity for paying customers, throttle the number of requests a customer can make in a given timeframe, or raise prices.</p> <p>In response, SaaS companies are already altering their billing structures. GitHub Copilot Pro, for example, has moved away from per-user pricing to a consumption-based model. That worries enterprise IT buyers who often lack the internal instrumentation to measure their token usage, leaving them unable to predict or monitor their expenditure until the monthly bills arrive.</p> </section> <section class="section main-article-chapter" data-menu-title="Measuring AI maturity"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Measuring AI maturity</h2> <p>Turning to how effectively organisations are deploying AI, Gartner vice-president Pieter den Hamer pointed to the analyst firm’s AI maturity model, which provides a comprehensive overview of the capabilities required to successfully scale AI.</p> <p>Chandrasekaran described six dimensions of this maturity. Organisations must focus on value by measuring return on investment and using it as a primary metric for use cases. They must ensure data quality, as <a href="https://www.techtarget.com/searchenterpriseai/tip/Exploring-the-context-layer-for-AI-systems">AI is only as useful as its contextual information</a>.</p> <p>They also need strict governance for risk mitigation, and strong engineering practices to ground systems in corporate data and automate deployments. Finally, businesses must balance top-down and bottom-up organisational innovation, while fostering a culture of psychological safety so employees feel empowered and secure in their long-term future.</p> <p>A global Gartner survey found that just 17% of companies are at a high level of maturity, having successfully scaled AI pervasively across their business functions. Another 51% sit at a medium level, generating some value in isolated pockets but largely struggling to demonstrate clear returns on investment. The remaining 32% exhibit low maturity, limiting their efforts to ideation, experiments or pilots.</p> </section> <section class="section main-article-chapter" data-menu-title="The importance of AI literacy"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The importance of AI literacy</h2> <p>While generative AI and traditional AI remain the top two investment areas, <a href="https://www.computerweekly.com/news/366639619/Emerging-markets-prioritise-top-line-growth-with-agentic-AI">agentic AI is showing significant growth</a>. Yet, den Hamer noted that only about one in five companies currently has AI agents in production, indicating a clear gap between supplier marketing hype and the practical reality of enterprise adoption. Companies are routinely stumbling over seemingly simple hurdles, such as identifying the right use cases and grappling with poor internal data quality, which starves AI implementations of essential context.</p> <p>Despite these challenges, clear best practices are emerging from the most mature organisations.</p> <p>Many businesses initially deploy AI to boost employee efficiency, but these productivity gains are often modest. Den Hamer also noted that, despite AI being used as a convenient scapegoat for recent technology sector restructuring, only a small percentage of industry layoffs can be directly attributed to AI.</p> <p>Instead of focusing purely on headcount and productivity, mature companies are using AI to accelerate research and development, enhance manufacturing quality, and improve overall business resilience. Furthermore, these organisations are embedding AI deeply into their business processes.</p> <p>Den Hamer warned that the vast majority of current AI initiatives are acting merely as a band-aid. Organisations need to <a href="https://www.computerweekly.com/news/366643396/Why-business-process-reinvention-is-needed-for-agentic-AI-workflows">rethink their workflows entirely</a> to maximise the technology’s potential, rather than accepting the over-optimistic narrative that AI is mature enough to fully replace human workers. The best recipe for success, Den Hamer noted, is seeking synergy between humans and AI, rather than trying to replace people completely with AI.</p> <p>Ultimately, that success hinges on educating people. Den Hamer concluded that there is a surprisingly strong correlation between financial returns and actively fostering AI literacy, noting that training and guiding staff at all levels of the organisation is key for them to gain an understanding of AI.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li><a href="https://www.computerweekly.com/news/366644858/OpenAI-deepens-Japan-footprint-with-Hitachi-deal">Hitachi will use OpenAI’s Codex agent</a>&nbsp;to unpick ageing mission-critical systems and gain early access to its frontier AI models in a slew of high-profile Japanese partnerships for the US AI lab.</li> <li>Oxford Economics report reveals that pursuing&nbsp;<a href="https://www.computerweekly.com/news/366644332/Strict-sovereign-AI-policies-could-cost-APAC-economies-billions">total AI self-sufficiency will lead to economic trade-offs</a>, delayed enterprise adoption and higher carbon footprints across the Asia-Pacific region.</li> <li>Alibaba Cloud unveils&nbsp;<a href="https://www.computerweekly.com/news/366643987/Alibaba-Cloud-opens-Johor-cloud-region">two new datacentres in Johor</a>, cementing its largest infrastructure presence in Southeast Asia while capitalising on spillover demand from Singapore.</li> <li><a href="https://www.computerweekly.com/news/366643677/Kmart-taps-Google-AI-to-launch-virtual-try-ons-in-retail-first">Kmart is deploying Google Cloud’s AI capabilities</a>&nbsp;to let customers preview clothes on themselves and visualise furniture in their homes as it embraces conversational commerce to win over shoppers.</li> </ul> </div> </div> </section> As foundational artificial intelligence capabilities converge, Gartner analysts urge IT leaders to focus on data quality, AI literacy and process integration, among other areas, rather than chasing the latest models https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/large-language-model-LLM-generative-AI-chat-gpt-peopleimagescom-adobe.jpg https://www.computerweekly.com/news/366645093/Gartner-warns-AI-model-advantage-is-shrinking Wed, 24 Jun 2026 17:00:00 GMT <p>Data and IT leaders are under pressure to deliver business outcomes from <a href="https://www.computerweekly.com/resources/Artificial-intelligence-automation-and-robotics">artificial intelligence</a> (AI) initiatives amid ongoing industry hype and fears of a bursting bubble, but achieving true business value goes beyond <a href="https://www.techtarget.com/searchcio/definition/ROI">return on investment</a>&nbsp;(ROI).</p> <p>That’s because AI represents more than just a technology shift, Gartner’s vice-president analyst, Jorg Heizenberg, noted at the research firm’s <a href="https://www.gartner.com/en/conferences/apac/data-analytics-australia/sessions/detail/4537142-Gartner-Opening-Keynote-Navigate-AI-on-Your-Data--Analytics-Journey-to-Value">Data and Analytics Summit keynote</a> in Sydney last week, adding that it marks a change that could be just as profound as the arrival of the internet.</p> <p>In navigating the AI era, Gartner director-analyst Georgia O’Callaghan noted that while nearly three in five organisations had put an AI service in production in 2025 and four in five are doubling down on AI today, “you can’t just continue to increase your investments in AI without getting clarity on the goals and ambition of your organisation”.</p> <p>Heizenberg warned that data and analytics professionals should redefine their AI ambitions with input from stakeholders, particularly regarding their tolerance for AI disruption. Those with a low tolerance can choose a cautious approach, carefully assessing risk and following the safest course. Those with a greater appetite for disruption can take a more opportunistic approach, while organisations with a high tolerance might dare to be pioneers, even if that means taking the biggest risks.</p> <p>However, one of the first questions stakeholders ask is: “What is this going to cost?” This is a difficult question to answer because <a href="https://www.computerweekly.com/news/366599472/How-to-stop-AI-costs-from-soaring">AI costs are highly unpredictable</a> and often hidden, O’Callaghan pointed out. The problem is compounded by vendors using pricing models based on metrics that are difficult to forecast, such as graphics processing unit (GPU) hours and <a href="https://www.computerweekly.com/news/366641232/Digital-Realty-CTO-on-AI-tokenomics-and-datacentre-infrastructure">token consumption</a>.</p> <p>Gartner’s research shows that while six out of 10 IT leaders are worried about AI agents running up unexpected costs, only two out of 10 data and AI leaders are concerned that unpredictable pricing might limit the value they get from the technology. This disconnect should be a wake-up call, said Heizenberg.</p> <p>“AI can be an expensive lesson,” warned O’Callaghan, noting that less than half of organisations manage and optimise their AI-related spending. Organisations should track expenses from the outset – especially during prototyping – and adopt cost-driven design to understand the financial impact of various components before going into production. For instance, teams should explore the <a href="https://www.techtarget.com/searchcio/feature/The-hidden-costs-of-AI-What-leaders-must-budget">cost implications of using different large language models</a>&nbsp;(LLMs), or even <a href="https://www.techtarget.com/whatis/definition/small-language-model-SLM">small language models</a>&nbsp;(SLMs), to power an AI agent.</p> <p>When communicating with stakeholders, however, the focus should remain on value rather than just cost – and there is more to value than money. Heizenberg highlighted North Yorkshire Council, which created a digital citizen named Dotty and mapped her journey through public services to make the impact of data relatable for all employees.</p> <p>For example, transposing two digits in a home address might result in a tradesperson being sent to the wrong house to install a handrail for an elderly person. That wasted journey carries a direct financial cost, but there are also ripple effects: what if the lack of a handrail results in the resident falling and suffering a serious injury?</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The change management and training effort for AI tools takes nearly twice as long as implementing the AI solution itself, which means planning for longer timelines and higher costs than for any other technology implementation you have managed before </figure> <figcaption> <strong>Georgia O’Callaghan, Gartner</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>Whatever an organisation’s ambition, foundational investments are key. A 2025 Gartner survey on modern data realisation found that respondents who were most satisfied with the outcomes of their AI use cases spent 30% more on foundational activities, such as data management, governance and talent, compared with those who were unsatisfied.</p> <p>Other Gartner surveys found 59% of IT leaders felt they were being pushed into adopting <a href="https://www.computerweekly.com/news/366612652/APAC-organisations-embrace-generative-AI">generative AI</a> (GenAI) tools before they were ready, while 61% felt pressure from senior leaders, directors or stakeholders to move forward with AI.</p> <p>One of the biggest issues is whether an organisation’s data is secure and well-governed enough to be opened up to further AI applications, including autonomous agents.</p> <p>“We need to prevent the exposure of the wrong data to the wrong people, applications or LLMs with AI governance, and avoid inaccuracies, misunderstandings and hallucinations with a well-designed context layer,” said O’Callaghan. “This will help to ensure that your data is AI-ready, trusted and aligned to the use case.”</p> <p>She added that this highlights the importance of repositioning governance as a business value accelerator, rather than a function focused purely on compliance.</p> <p>To improve AI governance, the analysts suggested three key steps. First, organisations should connect their existing governance groups, such as risk, data and cyber security, into a unified AI governance team. Gartner predicts organisations connecting governance bodies in this way will experience a 10% greater business impact than those that do not.</p> <p>The second step is to rationalise governance. This involves having the unified team review and consolidate various policies into a clear, consistent framework that reflects the organisation’s risk tolerance and cultural values around responsible AI use.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li><a href="https://www.computerweekly.com/news/366644858/OpenAI-deepens-Japan-footprint-with-Hitachi-deal">Hitachi will use OpenAI’s Codex agent</a>&nbsp;to unpick ageing mission-critical systems and gain early access to its frontier AI models in a slew of high-profile Japanese partnerships for the US AI lab.</li> <li>Oxford Economics report reveals that pursuing&nbsp;<a href="https://www.computerweekly.com/news/366644332/Strict-sovereign-AI-policies-could-cost-APAC-economies-billions">total AI self-sufficiency will lead to economic trade-offs</a>, delayed enterprise adoption and higher carbon footprints across the Asia-Pacific region.</li> <li>Alibaba Cloud unveils&nbsp;<a href="https://www.computerweekly.com/news/366643987/Alibaba-Cloud-opens-Johor-cloud-region">two new datacentres in Johor</a>, cementing its largest infrastructure presence in Southeast Asia while capitalising on spillover demand from Singapore.</li> <li><a href="https://www.computerweekly.com/news/366643677/Kmart-taps-Google-AI-to-launch-virtual-try-ons-in-retail-first">Kmart is deploying Google Cloud’s AI capabilities</a>&nbsp;to let customers preview clothes on themselves and visualise furniture in their homes as it embraces conversational commerce to win over shoppers.</li> </ul> </div> </div> <p>Finally, governance must be <a href="https://www.computerweekly.com/opinion/Navigating-culture-to-govern-AI-successfully">embedded across both the culture and the technology of the business</a>. This requires shifting the organisational mindset from compliance to one where everyone understands how to use data responsibly and ethically.</p> <p>Technologically, leaders should adopt <a href="https://www.computerweekly.com/news/366558852/Why-IT-governance-is-a-coding-issue">policy-as-code</a> so rules are automatically enforced throughout the tech stack. Gartner predicts that by 2028, organisations using specialised governance tools will decrease the cost of regulatory compliance by up to 20%.</p> <p>Even when data is well-governed, context remains critical. If an employee asks how many active customers the business has, the answer depends on the definition of “active”. Does it mean someone who made a recent purchase, holds an ongoing subscription, or recently visited the website?</p> <p>In the absence of context, an LLM can easily misunderstand the prompt and rapidly amplify that error.</p> <p>“It’s time to build an integrated <a href="https://www.techtarget.com/searchenterpriseai/tip/Exploring-the-context-layer-for-AI-systems">context realisation layer</a> – a layer that connects every piece of information so everyone and everything, people and agents alike, can see the bigger picture and make more informed decisions,” said Heizenberg.</p> <p>While semantic layers are becoming commonplace, they are no longer sufficient on their own. Organisations are now experimenting with ontologies, <a href="https://www.techtarget.com/searchdatamanagement/tip/5-knowledge-graph-use-cases-in-data-fabric-architecture">knowledge graphs</a>&nbsp;and other methods to attach deeper meaning to data. Combining these approaches yields far more accurate results.</p> <section class="section main-article-chapter" data-menu-title="Investing in people"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Investing in people</h2> <p>Another challenge is that technology is evolving faster than the workforce can adopt it. “If you’re investing in AI without investing in your people, you are throwing money away,” warned O’Callaghan. “The change management and training effort for AI tools takes nearly twice as long as implementing the AI solution itself, which means planning for longer timelines and higher costs than for any other technology implementation you have managed before.”</p> <p>To counter this, a “mindset, skillset, toolset” approach is highly effective. IT leaders must ask: “What mindset obstacles exist in the organisation, and how can they be overcome? What skills gaps are present, and how can they be remedied?” Only after addressing mindset and skillset should leaders ask what tooling changes are needed.</p> <p>Finally, there is the ongoing concern about AI-driven job losses. Gartner found that 34% of CIOs expect to reduce the size of their workforce over the next three years. Conversely, only 4% of chief data officers have decreased their team size in the past year, while 44% have expanded their teams.</p> <p>“Currently, we’re not seeing much reduction in data and analytics team size, but this is happening in other areas,” said O’Callaghan.</p> <p>She noted, however, that some organisations may be using the introduction of AI as a convenient excuse for layoffs that would have occurred regardless.</p> <p>“The value of human skills and talent will still sit at the core of delivery teams, but these teams will now combine human expertise with AI agents to make more productive, AI-powered fusion teams,” said O’Callaghan.</p> </section> Gartner analysts call for IT leaders to prioritise foundational investments in governance, change management and talent to realise the benefits of AI https://cdn.ttgtmedia.com/visuals/German/Hero-Ai-Ki-Leo-Lintang-Adobe-05.jpg https://www.computerweekly.com/news/366645092/Gartner-Prioritise-governance-to-beat-AI-hype Tue, 23 Jun 2026 21:34:00 GMT <p>Australia’s whole-of-government cloud policy comes into effect on 1 July 2026, establishing cloud as the default when modernising IT infrastructure. The policy document, prepared by the Digital Transformation Agency (DTA), sets out five broad requirements.</p> <p>They include the need for government entities to prioritise cloud technologies when <a href="https://www.techtarget.com/searchenterpriseai/tip/Integrate-and-modernize-legacy-systems-with-AI">modernising IT infrastructure</a>; tap cloud technologies to drive innovation, including <a href="https://www.computerweekly.com/resources/Artificial-intelligence-automation-and-robotics">artificial intelligence (AI)</a>; adopt cloud securely and responsibly; actively manage and optimise cloud computing costs; and nurture cloud skills across the Australian Public Service (APS).</p> <p>The first specific requirement of the modernisation policy is for agencies to adopt cloud solutions for all new digital and ICT initiatives and upgrades unless an alternative is justified.</p> <p>However, Gartner director-analyst Adrian Wong warned that a blanket mandate overlooks the reality that an application or workload component may simply be a poor fit for a cloud solution. Legacy applications, for example, often fail to fully utilise cloud computing capabilities. This makes them technically mismatched and sometimes unexpectedly more expensive to run in the cloud than in a local datacentre.</p> <p>There is cause for concern. Wong pointed out that while the policy frames this as a transition away from ageing systems rather than a strict requirement to migrate every existing legacy app to the cloud, aggressive timelines can drive poor decision-making.</p> <p>If organisations feel rushed, especially if they lack adequate cloud planning and architectural expertise, they are more likely to pursue poorly conceived <a href="https://www.computerweekly.com/feature/APAC-expert-guide-to-cloud-migration">lift-and-shift migrations</a>. These hurried efforts frequently fail to meet expectations and form the basis for cloud project failures.</p> <p>The reasons for such failures, according to a Gartner report on handling cloud project failures, include workloads being inappropriate for the cloud, poorly chosen providers, bad design or implementation, inaccurate cost estimates and integration issues.</p> <p>Some factors make workloads inherently more suited to on-premise deployment, Wong noted. These include high sensitivity to latency; strict data residency, compliance or sovereignty mandates that cannot be satisfied with public cloud solutions; unique service-level agreements that cloud providers might not be able to meet; and environments requiring enterprise-controlled assets.</p> <p>“Ultimately, avoiding cloud dissatisfaction requires agencies to have the time and flexibility to perform a detailed application portfolio analysis. While prioritising modern cloud solutions is a strong strategic aspiration, enforcing rigid decommissioning pressures risks forcing bad long-term fits just to satisfy policy requirements,” Wong warned.</p> <p>Vinayak Sreedhar, country manager for Australia and New Zealand (ANZ) at ManageEngine, an IT management and monitoring provider that serves federal, state and local government customers, said agencies shouldn’t underestimate the complexity of what lies ahead and that migrating away from legacy systems while ensuring ongoing compliance is no easy feat.</p> <p>“The agencies most at risk are those without a clear picture of what’s being retired, when, and what is dependent on it,” Sreedhar said. “Moving fast without that clarity is how outages occur.”</p> <section class="section main-article-chapter" data-menu-title="AI and interoperability"> <h2 class="section-title"><i class="icon" data-icon="1"></i>AI and interoperability</h2> <p>Cloud platforms are seen as a way of creating a more connected, responsive and data-driven public sector, in part through the adoption of artificial intelligence (AI).</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> While prioritising modern cloud solutions is a strong strategic aspiration, enforcing rigid decommissioning pressures risks forcing bad long-term fits just to satisfy policy requirements </figure> <figcaption> <strong>Adrian Wong, Gartner</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>While government entities are required to design for interoperability and portability to minimise supplier lock-in, they are only encouraged to ensure cloud services support open standards <a href="https://www.computerweekly.com/feature/Enterprise-strategies-for-API-management">and application programming interfaces (APIs)</a>, and allow for data portability.</p> <p>SUSE ANZ general manager Ben Henshaw suggested the language in the policy gives the impression that the DTA wants to avoid another “mother of all lock-in” situation that repeats <a href="https://www.computerweekly.com/news/252473040/Businesses-at-risk-as-mainframe-skills-die-out">historical problems with mainframes</a>. Once data is locked into a particular cloud, it becomes very hard and costly to extract it into a format that can be deployed elsewhere.</p> <p>Public clouds are designed as a “land grab” to capture as many departmental workloads as possible, Henshaw warned. “They’re not making it easy to get out because why would they? It’s not in their commercial interest to be open, interoperable, more standard spaces.” For example, hyperscalers each have their own domain-specific languages for creating templates that specify operating systems and software for virtual machines.</p> <p>The government cloud policy highlights the design and procurement principles of selecting architectures that are open, interoperable, contestable and portable – ideally without needing to hire a thousand consultants for a replatforming exercise – but that remains a challenge, he suggested.</p> <p>Part of the problem for governments and businesses alike is that a vast amount of money is spent simply keeping the lights on and upgrading, rather than on innovation. Replatforming with low cost and effort is the “secret sauce” of open source, and of companies like SUSE, because they are agnostic, Henshaw said. This allows agencies to spend more time deploying new features rather than draining budgets on system upgrades.</p> <p>While SUSE’s cloud provider partners offer utility, Henshaw admitted they also pose risks and add cost because they rely on proprietary technology stacks, creating complications for <a href="https://www.computerweekly.com/feature/A-better-way-to-manage-hybrid-or-multicloud-deployments">multicloud environments</a>. Departments such as education, health, defence, home affairs and Services Australia are complex organisations with vast use cases, and cannot source all their capabilities from a single provider like Amazon Web Services, Google Cloud, Microsoft Azure, Oracle or SAP. This makes interoperability, portability and integration vital.</p> <p>Agentic AI is also gaining attention as a way to automate workflows. Different systems within a process will use different <a href="https://www.computerweekly.com/news/366638829/Large-language-models-provide-unreliable-answers-about-public-services-Open-Data-Institute-finds">large language models (LLMs)</a> of varying sizes, meaning data processing needs will be highly varied. At one extreme, soldiers have disconnected, intermittent and limited (DIL) access to remote systems, meaning processing must be done locally. At the other extreme, the health department processes large volumes of records to determine benefits or treatments.</p> <p>With many LLMs available, both open source and proprietary, Henshaw said it is incredibly important for governments to retain sovereign control over their data and models. Governments are looking to open source LLMs to access the code, ensure explainability and govern the models.</p> <p>According to Sreedhar, the explicit push to embed AI readiness across cloud platforms is forward thinking and necessary, but it isn’t a switch organisations can simply flip post-migration.</p> <p>“How is the data structured, governed and stored? How much compute is being provisioned? And how will models eventually be deployed? These questions require deliberate architectural decisions from day one,” Sreedhar said. “Those that treat AI as a future add-on rather than a current design requirement will be hit with expensive infrastructure rebuilds in a few years’ time. The time to get this right is during the transition, not after.”</p> </section> <section class="section main-article-chapter" data-menu-title="Security considerations"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Security considerations</h2> <p>Henshaw pointed out that federal government agencies will have to navigate the cloud transition whether it proves hard or easy, especially when it comes to security.</p> <p>“No one wants to be on the front page of the newspaper. Nobody wants to be the person who accidentally put information out into a public AI system that caused a whole lot of sovereign angst,” he said. A modern, defensible architecture is an essential, non-negotiable requirement for hosting and running AI workloads safely and securely.</p> <p>As a supplier, part of SUSE’s job is to help government departments apply a modern defensible architecture, adhering to <a href="https://www.computerweekly.com/news/252478146/What-should-be-in-Australias-next-cyber-security-strategy">Essential Eight principles</a>, the Australian Signals Directorate’s information security manual and ISO 27001. This ensures a <a href="https://www.computerweekly.com/opinion/Zero-trust-is-redefining-cyber-security-in-2025">zero-trust architecture</a> that is portable, composable and interoperable. Without this, Henshaw suggested, federal agencies will lag in their ability to tap the technical benefits of AI.</p> <p>Sreedhar warned that the sheer scale of the transition creates a much larger attack surface. <a href="https://www.computerweekly.com/news/366613412/Australia-bolsters-cyber-defences-with-security-bill">Recent cyber security legislative reforms</a> have sharpened obligations for critical infrastructure operators to protect business-critical data, but agencies should treat those obligations as a mere baseline.</p> <p>“The vulnerability we see most often in cloud transitions isn’t technical – it’s the gap between IT teams and security teams during the migration itself,” Sreedhar said. “Security architects need to be part of the transition from procurement through to go-live and beyond.”</p> </section> <section class="section main-article-chapter" data-menu-title="Skills uplift"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Skills uplift</h2> <p>A policy framework is only as good as the people who put it into practice, Sreedhar observed. The DTA has been clear that agencies must build the skills, infrastructure and governance required to meet community expectations, yet workforce capability is almost always the most underfunded component of digital transformation.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Getting the technology right matters, but so does building a public service that understands and owns what it’s building </figure> <figcaption> <strong>Vinayak Sreedhar, ManageEngine</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>Agencies should be evaluating their internal capability right now, well ahead of the 1 July deadline, and <a href="https://www.computerweekly.com/opinion/Gartner-How-to-overcome-a-lack-of-cloud-skills-in-your-organisation">investing in genuine skills uplift</a> where gaps exist.</p> <p>“Getting the technology right matters, but so does building a public service that understands and owns what it’s building,” Sreedhar said. This is especially vital for the policy’s fifth requirement, which explicitly demands agencies nurture cloud skills across the APS.</p> <p>“Agencies won’t be able to satisfy the policy simply by pointing to cloud deployments. That’s the easy part,” Sreedhar continued. “Agencies need genuine workforce development strategies and plans to close identified skills gaps. One of the ways we’re addressing this at ManageEngine is at the operational layer, helping staff build fluency with hands-on training and tools spanning infrastructure, security and <a href="https://www.computerweekly.com/news/366641816/How-the-AI-boom-is-reshaping-tech-cost-management">FinOps</a> – the disciplines the DTA has specifically and rightly called out.”</p> <p>Reflecting on the skills mandate, Henshaw described this aspect of the policy as a strong starting point that offers good principles and guidelines. “It’s there not as a stick, but as a compass,” he said.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about IT in Australia</h3> <ul class="default-list"> <li>The <a href="https://www.computerweekly.com/news/366643682/How-Canberra-Institute-of-Technology-is-transforming-classroom-learning">Canberra Institute of Technology partnered with Cisco</a> to standardise the institute’s physical and virtual classrooms, boosting inclusivity and slashing on-site support</li> <li>Melbourne-based Heidi is building its own AI models and launching wearable hardware to automate documentation and&nbsp;<a href="https://www.computerweekly.com/news/366640992/Aussie-AI-health-tech-Heidi-aims-to-cure-clinical-burnout">reduce the administrative burden on doctors</a>.</li> <li>ANZ Bank has started&nbsp;<a href="https://www.computerweekly.com/news/366638802/ANZ-rolls-out-AI-agents-for-business-bankers">rolling out AI agents within its new CRM system</a>&nbsp;to help business bankers recover hours of lost productivity by automating tasks and streamlining workflows.</li> <li>Oracle has&nbsp;<a href="https://www.computerweekly.com/news/366640566/Oracle-opens-Sydney-customer-excellence-centre-to-boost-AI-adoption">opened an AI customer excellence centre in Sydney</a>&nbsp;to help its customers across Australia and Oceania adopt the technology.</li> </ul> </div> </div> </section> As Australia prepares to enforce its whole-of-government cloud policy, industry experts warn agencies against rushed migrations, supplier lock-in and treating AI readiness as an afterthought https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/Australia-government-parliament-fotolia.jpg https://www.computerweekly.com/news/366644964/Australian-government-cloud-mandate-sparks-migration-warnings Tue, 23 Jun 2026 03:27:00 GMT <p>Few IT executives feel the pace of developments in <a href="https://www.computerweekly.com/resources/Artificial-intelligence-automation-and-robotics">artificial intelligence</a> (AI) as acutely as Manu Narayan. Some nine months into his role as the first chief information officer (CIO) at GitLab – the software development platform with over $1bn in revenue and more than 2,000 employees – Narayan is tasked with turning the company into a proving ground for the very technologies its customers use.</p> <p>“The AI space in general is changing so rapidly that we’ve constantly had to revisit our goals and things that we want to accomplish,” he said in a recent interview with Computer Weekly.</p> <p>With product development sitting with GitLab’s research and development team, Narayan’s mandate is mostly internal: modernising the business application stack, user support, as well as data and analytics. But instead of bolting AI onto existing workflows, his goal is to rebuild operations from the ground up.</p> <p>“When I was revisiting our AI strategy a few months ago, the focus was not on how we introduce AI,” he said. “The focus was to rethink the nature of work internally, leveraging AI. It’s thinking about processes from first principles and then using agentic AI to drive them.”</p> <p>Pointing to a customer success manager (CSM) as an example, Narayan noted that the purpose of the role is to build deep client relationships, yet CSMs spend hours on administrative tasks such as building quarterly business review slides for clients, transcribing notes and hunting for context across customer relationship management systems, data warehouses and chat channels.</p> <p>By deploying AI agents to handle the grunt work, GitLab is looking to free up its workforce to focus on high-level strategy. “We want all of our team members to focus on what matters most: the core purpose of their role,” said Narayan. “We’re leveraging AI for tasks that can help them scale out in a more linear way, more than just a 10-15% increase in productivity.”</p> <p>To manage AI deployments, GitLab has adopted a hub-and-spoke operating model. A central AI enterprise team handles governance, technical building and guardrails, while dedicated “AI transformation owners” embedded in individual divisions identify time-consuming, repeatable work that is ripe for automation.</p> <p>The approach has already been applied to GitLab’s own internal employee support network. The company has built AI agents to assist its 120 internal support staff across IT, people operations and sales, helping them instantly pull the context they need or deflect routine tickets entirely.</p> <section class="section main-article-chapter" data-menu-title="Rejecting ‘tokenmaxxing’"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Rejecting ‘tokenmaxxing’</h2> <p>As AI adoption increases across the enterprise, CIOs will naturally grapple with <a href="https://www.computerweekly.com/news/366641816/How-the-AI-boom-is-reshaping-tech-cost-management">cost control and measurement</a>. However, Narayan is wary of strategies such as “<a href="https://www.techtarget.com/searchcio/feature/Tokenmaxxing-How-CIOs-extract-maximum-value-AI-tokens">tokenmaxxing</a>”, where developers and employees are encouraged to maximise the number of AI tokens they use.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> It’s easy to get to 90% of an application you develop in-house. That last 10% – the role-based access controls, auditability, immutable logging, which are things you need as a public company or as a company that deals with regulated customers – is incredibly complex to build </figure> <figcaption> <strong>Manu Narayan, GitLab</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>“We’ve specifically avoided and don’t want to do tokenmaxxing,” said Narayan. “Gamification can help drive outcomes, but I think it drives the incorrect behaviour. We’re not looking for purely context-in, context-out as the measure of success. It’s really hard to know if somebody’s gaming the system. Are they just sending excessive content because they don’t actually know what they’re doing?”</p> <p>Instead of tracking token burn, GitLab tracks daily active usage across the tech stack to ensure its workforce is building sustainable habits. For calculating hard return on investment (ROI), Narayan insists on anchoring AI deployments to traditional business metrics. For an AI agent assisting a sales development representative, success isn’t measured by the number of prompts generated, but by standard key performance indicators: outbound messages, meetings scheduled and sales pipeline conversion.</p> </section> <section class="section main-article-chapter" data-menu-title="Build vs buy and the future of SaaS"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Build vs buy and the future of SaaS</h2> <p>As AI lowers the barrier to building internal tools, there have been suggestions that the <a href="https://www.techtarget.com/searchitoperations/news/366639662/SaaSpocalypse-Maybe-not-but-SaaS-applications-are-changing">days of off-the-shelf software-as-a-service (SaaS) applications are numbered</a>. Narayan views this as vastly overstated, particularly from a governance and compliance perspective.</p> <p>“We may see more custom interfaces and the disaggregation of systems of interaction from systems of record,” he said. “But the underlying governance controls in core SaaS tools aren’t going anywhere.”</p> <p>Narayan also pointed to the hidden costs of bespoke software development: “It’s easy to get to 90% of an application you develop in-house. That last 10% – the role-based access controls, auditability, immutable logging, which are things you need as a public company or as a company that deals with regulated customers – is incredibly complex to build.”</p> <p>To ensure safety across custom and supplier tools, GitLab grounds its AI governance in a strict data classification standard. Public data flows through self-service platforms, while proprietary or customer data requires deeper security reviews before interacting with <a href="https://www.computerweekly.com/feature/LLMs-explained-A-developers-guide-to-getting-started">large language models</a>.</p> <p>Despite strong executive backing and budget, change management remains a challenge for Narayan. Bridging the gap between AI-forward employees and those who are slower to adapt requires a mix of departmental centres of excellence and <a href="https://www.computerweekly.com/news/366644461/How-AngelHack-uses-hackathons-to-ease-AI-adoption">internal AI hackathons</a>.</p> <p>Yet, for a CIO, the greatest pressure is the ticking clock.</p> <p>“The thing that keeps me up at night is whether we’re moving fast enough,” said Narayan. “In the AI era, our decision-making needs to happen in days and weeks, not months and quarters. But I still worry about whether we are driving the right initiatives that are going to have the right long-term ROI for us.”</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li><a href="https://www.computerweekly.com/news/366644858/OpenAI-deepens-Japan-footprint-with-Hitachi-deal">Hitachi will use OpenAI’s Codex agent</a> to unpick ageing mission-critical systems and gain early access to its frontier AI models in a slew of high-profile Japanese partnerships for the US AI lab.</li> <li>Oxford Economics report reveals that pursuing <a href="https://www.computerweekly.com/news/366644332/Strict-sovereign-AI-policies-could-cost-APAC-economies-billions">total AI self-sufficiency will lead to economic trade-offs</a>, delayed enterprise adoption and higher carbon footprints across the Asia-Pacific region.</li> <li>Alibaba Cloud unveils <a href="https://www.computerweekly.com/news/366643987/Alibaba-Cloud-opens-Johor-cloud-region">two new datacentres in Johor</a>, cementing its largest infrastructure presence in Southeast Asia while capitalising on spillover demand from Singapore.</li> <li><a href="https://www.computerweekly.com/news/366643677/Kmart-taps-Google-AI-to-launch-virtual-try-ons-in-retail-first">Kmart is deploying Google Cloud’s AI capabilities</a> to let customers preview clothes on themselves and visualise furniture in their homes as it embraces conversational commerce to win over shoppers.</li> </ul> </div> </div> </section> Manu Narayan tells Computer Weekly why he’s steering clear of vanity metrics such as ‘tokenmaxxing’, why reports of SaaS’s death are overblown, and why the biggest pressure is simply keeping pace https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/code-metadata-programming-maciek905-adobe.jpg https://www.computerweekly.com/news/366644996/GitLab-CIO-rejects-tokenmaxxing-as-it-rebuilds-work-around-agentic-AI Mon, 22 Jun 2026 04:45:00 GMT <p>More than just delivering faster access speeds, the <a href="https://www.computerweekly.com/feature/6G-networks-explained-everything-you-need-to-know">transition to 6G mobile networks</a> will require the telecom industry to support the growing volume of artificial intelligence (AI) traffic, manage the integration of non-terrestrial networks (NTNs), and – more importantly – crack the enterprise market.</p> <p>To understand the roadmap to 2030, <a href="https://www.computerweekly.com/news/366644794/Uplink-traffic-gains-momentum-as-5G-subscriptions-top-three-billion">when the first commercial 6G networks are expected to be deployed</a>, Computer Weekly sat down with Balaji Raghothaman, chief technologist for 6G at Keysight Technologies, on the sidelines of the <a href="https://www.3gpp.org/news-events/3gpp-news/tsg112">3GPP’s first-ever plenary meeting in Singapore</a> last week. Given Keysight’s extensive footprint in network testing and measurement, Raghothaman has a front-row seat to the innovations and struggles of chipmakers, mobile operators and hyperscalers alike.</p> <p>In a wide-ranging interview, Raghothaman warned that telecom operators must rethink their business models to avoid becoming “dumb pipes” for AI traffic. He also discusses the industry’s move away from sub-terahertz spectrum, the need for space-based datacentres and why high-fidelity indoor digital twins will be the mobile industry’s ultimate weapon against Wi-Fi.</p> <p><i>Editor’s note: This interview has been edited for clarity and brevity.</i></p> <section class="section main-article-chapter" data-menu-title="As the industry progresses towards 6G, there are significant technical, business, and even geopolitical challenges that need to be overcome. What will it take for the industry to get there?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>As the industry progresses towards 6G, there are significant technical, business, and even geopolitical challenges that need to be overcome. What will it take for the industry to get there?</h2> <p><b>Balaji Raghothaman:</b> Fundamentally, <a href="https://www.computerweekly.com/opinion/Why-AI-monetisation-not-6G-is-the-real-prize-for-telcos">AI has been a huge inflection</a>. Whether we like it or not, it’s happening, and AI traffic dominates everywhere. It is inevitable that it’s happening for wireless traffic as well. Today, AI traffic originates from phones, but five or seven years from now, you’d have smart glasses, body sensors and robots. That will fundamentally change what the network is used for and there’s both opportunity and danger in that.</p> <p>In 4G and the early days of 5G, there was the rise of the app economy, and somehow telcos ended up as just the pipe while others took value. The same thing may happen with 6G if telcos are not careful – that they’re just the pipe delivering AI traffic, and value is captured by someone else. Some forward-thinking operators are looking at concepts such as <a href="https://www.techtarget.com/searchnetworking/definition/What-is-artificial-intelligence-radio-access-network-AI-RAN">AI-RAN [radio access network]</a>, where they share server resources to deliver RAN traffic and run AI workloads, creating new business cases.</p> </section> <section class="section main-article-chapter" data-menu-title="From a technical perspective, there has been a lot of talk about using terahertz bandwidth for 6G, but those frequencies have very short propagation distances. What are the implications for infrastructure build-out?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>From a technical perspective, there has been a lot of talk about using terahertz bandwidth for 6G, but those frequencies have very short propagation distances. What are the implications for infrastructure build-out?</h2> <p><b>Raghothaman</b>: Early on, when people started talking about 6G, sub-terahertz was a big topic. There was a lot of investigation, but as you said, the range is very limited, and they are expensive to deploy. As we’ve come closer to 6G becoming a reality, the interest has dwindled. There’s very little appetite – and no discussion in 3GPP – right now on sub-terahertz. The focus is generally on existing frequency range [FR] bands: FR1, FR2 and FR3.</p> <p>Globally, there is consensus among operators that they would like to deploy 6G with the same footprint as FR1 bands – like the existing 5G C-band [around 3.3 GHz to 4.2 GHz] – on FR3 [about 7 GHz to 24 GHz] bands. It’s a doubling of frequency, so the propagation loss is higher and coverage is lower.</p> <p>The only way to reclaim coverage parity with FR1 is to use more antennas to create beamforming gain. In 5G, we talk about a maximum of 64 or 128 antennas. FR3 will most likely have 256, 512 or more antennas. That’s a big technology challenge. You have to design new radios without going over power budgets. There’s a danger that the energy-per-bit goes up when deploying on FR3, so creating more efficient power amplifiers and RF [radio frequency] front-ends is a major area of innovation.</p> </section> <section class="section main-article-chapter" data-menu-title="How will these spectrum and antenna challenges impact end-user devices, particularly with constraints on battery life and physical space?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>How will these spectrum and antenna challenges impact end-user devices, particularly with constraints on battery life and physical space?</h2> <p><b>Raghothaman:</b> The device at the edge still needs to reach the base station, which makes for an incredibly challenging design. Device makers can’t just remove old radios; they have to maintain legacy support while adding new ones. Real estate on phones is already limited.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Traditionally, you certify a device based on static behaviour. But if an AI-driven device changes its behaviour over time, how do you certify it? If it’s a drone equipped with a 6G modem flying around, how is that regulated? </figure> <figcaption> <strong>Balaji Raghothaman, Keysight</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>If you think a phone is challenging, think about smart glasses. There is virtually no real estate. We talk to engineers at companies like Meta, and they have to make decisions on every nanometre of space – whether to use it for battery power, processing, or transmitting. It’s a good problem to have because it drives innovation, but it is a major engineering hurdle.</p> </section> <section class="section main-article-chapter" data-menu-title="With 6G, there’s also a push to integrate the technology with NTNs like satellite communications. What are the interoperability challenges?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>With 6G, there’s also a push to integrate the technology with NTNs like satellite communications. What are the interoperability challenges?</h2> <p><b>Raghothaman:</b> First, there are spectrum coexistence issues. There are areas where NTN coincides with terrestrial spectrum, and decisions have to be made on how they will coexist – you cannot necessarily shut off the terrestrial base station when a satellite flies over.</p> <p>The other big challenge is about speed. A satellite moves at thousands of kilometres per hour, whereas a person is moving at three kilometres per hour relative to a terrestrial base station. This discrepancy, along with massive delays and Doppler effects, creates a big strain on the baseband receiver.</p> <p>But satellites are the only way you are going to get close to 100% coverage. In areas such as the vast jungles of Indonesia or Malaysia, it’s simply not economically viable for a commercial operator to maintain a terrestrial tower.</p> </section> <section class="section main-article-chapter" data-menu-title="And will we see a greater need for space-based datacentres with 6G?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>And will we see a greater need for space-based datacentres with 6G?</h2> <p><b>Raghothaman:</b> Satellite has two different kinds of implementations: one is pass-through which only relays signals, and the other is regenerative, which processes data. Because satellite delays are inherently long, you want to do AI inferencing as close to the edge as possible to deliver a good user experience. Putting a datacentre on a regenerative satellite makes sense. The real question is maintenance. You can climb a cell tower to fix a terrestrial antenna, but you can’t easily send a shuttle to space. This is why testing and validation are important to monitor satellites and predict when they might fail so a replacement launch can be scheduled.</p> </section> <section class="section main-article-chapter" data-menu-title="Let's talk about enterprise use cases. What sorts of enterprise applications will 6G enable that might finally move the needle?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Let's talk about enterprise use cases. What sorts of enterprise applications will 6G enable that might finally move the needle?</h2> <p><b>Raghothaman:</b> With every generation of mobile technology, there’s a big push towards enterprise solutions, and in the end, enterprises figure out that Wi-Fi is just easy. Deployment ease is a big issue. For IT managers, Wi-Fi is plug-and-play. 3GPP base stations are not like that.</p> <p>The other issue is that cellular network planning has traditionally been very outdoor-focused, hoping the signal penetrates indoors. But now, we finally have the <a href="https://www.keysight.com/blogs/en/inds/2025/3/creating-a-network-digital-twin-for-6g">tools available with ray tracing and digital twins</a>. Because we have building plans and material information, every building can have a geometric digital twin.</p> <p>Once you have that digital twin, you can use AI-powered network optimisation to arrive at a highly optimised, site-specific network design. The network can learn based on local usage patterns and dynamically tweak parameters. If the cellular guys do not figure out how to do indoor networks properly, they will lose out on a big portion of the enterprise market to Wi-Fi.</p> </section> <section class="section main-article-chapter" data-menu-title="What about network sensing? Could that drive new enterprise value, such as with self-driving vehicles in logistics or robotics?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>What about network sensing? Could that drive new enterprise value, such as with self-driving vehicles in logistics or robotics?</h2> <p><b>Raghothaman:</b> <a href="https://www.computerweekly.com/news/366622918/ETSI-highlights-integrated-sensing-communication-6G-use-cases">Integrated sensing and communication</a> is a big focus for 3GPP, but we have to be very careful. With 5G, we created all these great expectations of new use cases. I remember <a href="https://www.computerweekly.com/news/252492039/5G-IoT-technology-to-transform-health-and-social-care">remote surgery was a major use case</a>, but it didn’t really happen. We don’t want sensing to become that thing that fails and causes people to lose faith in the technology.</p> <p>Sensing is full of hard technical problems. Depending on the frequency you use, what is the resolution? Can you tell the difference between a bird and a drone? Those things still need to be proven. But everyone is excited about it because operators are looking for new ways to monetise beyond flat-rate subscriptions. If you can provide a service to a building owner or city government – whether that’s drone detection, elder care fall detection, or sensing a sag in a bridge – that is a whole new revenue stream.</p> </section> <section class="section main-article-chapter" data-menu-title="Finally, what does the regulatory landscape look like? When you mix AI, 6G, and autonomous devices, how are regulators adapting?"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Finally, what does the regulatory landscape look like? When you mix AI, 6G, and autonomous devices, how are regulators adapting?</h2> <p><b>Raghothaman</b>: Traditional regulatory work around spectrum harmonisation and terrestrial/satellite coexistence is ongoing. We are working hard to ensure we don’t repeat the issues where <a href="https://spectrum.ieee.org/faa-5g">5G deployments interfered with older aircraft radar altimeters</a>.</p> <p>But AI introduces entirely new regulatory questions. I represent Keysight on the FCC’s [Federal Communications Commission] technology advisory council in the US, where we are looking at what AI means for wireless networks. Traditionally, you certify a device based on static behaviour. But if an AI-driven device changes its behaviour over time, how do you certify it? If it’s a drone equipped with a 6G modem flying around, how is that regulated?</p> <p>The appetite for regulation varies – lower in the US, higher in Europe – but AI regulation is inevitably going to collide with telecom regulation in the 6G era. Our role at Keysight is to provide the measurement, data, and evidence so the industry can safely navigate wherever that landscape goes.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about 6G in APAC</h3> <ul class="default-list"> <li>SK Telecom has published a <a href="https://www.computerweekly.com/news/366639378/SK-Telecom-outlines-mid-to-long-term-6G-network-evolution">whitepaper in preparation for the commercialisation of 6G</a>, which it expects after 2030, emphasising the direction of communications infrastructure in the AI era.</li> <li>NTT Docomo accelerates <a href="https://www.computerweekly.com/news/366634587/NTT-DOCOMO-claims-successful-outdoor-6G-AI-driven-interface-trial">trial of 6G mobile technologies</a>, resulting in throughput improving by ‘up to 100%’ under real-world outdoor conditions.</li> <li><a href="https://www.computerweekly.com/news/366643457/Ericsson-Telstra-team-for-Australian-6G-development">Telstra and Ericsson join forces on 6G development</a> work spanning research, standards and real-world testing looking to pave way for the next era of advanced connectivity.</li> <li>Indosat has become the <a href="https://www.computerweekly.com/news/366620460/Indosat-Ooredoo-Hutchison-claims-AI-RAN-first-for-Southeast-Asia">first operator in Southeast Asia to build a commercial AI-RAN network</a>, enabling the convergence of AI and wireless connectivity to enhance performance and efficiency.</li> </ul> </div> </div> </section> Balaji Raghothaman, Keysight’s chief technologist for 6G, discusses what it takes for the telecoms industry to move to 6G, from supporting AI workloads to integrating with satellites connectivity and driving enterprise use cases https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/6g-mobile-network-broadband-sitthiphong-adobe.jpg https://www.computerweekly.com/news/366644835/Keysights-chief-technologist-on-the-path-to-6G Fri, 19 Jun 2026 03:02:00 GMT <p>Palo Alto Networks has billed the 2026 FIFA World Cup as the “largest global entertainment attack surface in history”, according to research from its Unit 42 threat intelligence and incident response arm.</p> <p>With the <a href="https://www.computerweekly.com/news/366634402/Lenovo-to-power-FIFA-World-Cup-2026">expanded 48-team tournament underway</a> across 16 host cities in the US, Canada and Mexico, billions of fans and a network of suppliers have entered the crosshairs of financially motivated cyber criminals, hacktivists and nation-state actors seeking disruption at scale.</p> <p>According to Unit 42 researchers, the logistical scale of the 39-day event, spanning four time zones and multiple regulatory regimes, is creating <a href="https://www.computerweekly.com/opinion/Why-asset-visibility-matters-in-industrial-cybersecurity">operational technology (OT) and IT security blind spots</a>.</p> <p>“Each match operates a layered, ring-based tournament network grafted onto a permanent stadium environment, depends on a temporary commercial supplier ecosystem, and pulls on host-city public services that FIFA does not own,” the Unit 42 report stated, noting that the reliance on fragmented, municipal infrastructure has vastly expanded the scope of potential targets for threat actors.</p> <section class="section main-article-chapter" data-menu-title="Profit, disruption and disinformation"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Profit, disruption and disinformation</h2> <p>The threat landscape surrounding the tournament has been categorised into three primary attack motives: disruption, profit and disinformation.</p> <p>While state-sponsored disinformation and disruptive attacks, such as <a href="https://www.computerweekly.com/news/366573272/More-DDoS-attacks-launched-against-APAC-financial-firms">distributed denial of service (DDoS) campaigns</a> and website defacements, are significant concerns, Palo Alto Networks noted that financially motivated cyber crime remains the “highest-volume, highest-likelihood threat”. Hackers have heavily industrialised their attacks against the hospitality sector since 2023, setting the stage for targeted hospitality ransomware affecting reservations, point-of-sale (POS) systems, and widespread fan fraud.</p> <p>However, the global geopolitical climate adds a layer of risk to the host nations’ critical infrastructure. Unit 42’s research noted that the <a href="https://www.computerweekly.com/news/366639621/Resilience-under-pressure-How-regional-conflict-is-reshaping-the-Middle-East-tech-strategy">recent conflicts in the Middle East</a> have reordered the threat surface for any US-hosted event.</p> <p>Researchers pointed to Iran-nexus threat groups, such as the Handala Hack Team and the Islamic Revolutionary Guard Corps (IRGC)-affiliated CyberAv3ngers, which have previously targeted internet-exposed industrial control systems.</p> <p>With the 2024 US Cybersecurity and Infrastructure Security Agency (CISA) assessment finding that over 70% of US water utilities are non-compliant with existing safety requirements, municipal water and energy grids in World Cup host cities remain highly lucrative targets for disruption.</p> <p>As the tournament moves from preparation to live operations, the window for threat mitigation is closing fast. Unit 42 is urging cyber defenders across the event’s supply chain to map out risks across the entire host-city ecosystem, <a href="https://www.computerweekly.com/opinion/Incident-response-planning-requires-constant-testing">stress-test their incident response plans</a> against realistic scenarios and ensure coordination across jurisdictions.</p> <p>History shows that where a strong security posture exists, mega-events operate without significant disruption; where defences are weak, adversaries succeed. Summarising the necessary mindset for security leaders, the researchers warned: “The single most important defender posture for 2026 is to assume the attacks will come.”</p> </section> <section class="section main-article-chapter" data-menu-title="Protecting fans in APAC"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Protecting fans in APAC</h2> <p>The cyber threat extends well beyond enterprise networks and municipal grids, directly targeting the estimated five to six million in-venue spectators and billions watching at home.</p> <p>In a media statement issued from Singapore today, Palo Alto Networks warned football fans across the Asia-Pacific region to maintain strong cyber hygiene. Cyber criminals are actively leveraging the World Cup fervour to push fake merchandise stores, fraudulent streaming platforms and <a href="https://www.techtarget.com/searchsecurity/feature/Quishing-on-the-rise-How-to-prevent-QR-code-phishing">malicious QR codes</a> at local viewing parties.</p> <p>To mitigate these consumer threats, Unit 42 advised fans to stick exclusively to FIFA-licensed platforms for streaming, warning against third-party sites, Telegram channels, and peer-to-peer payment apps offering free viewing.</p> <p>When booking accommodation or buying merchandise, fans should cross-reference listing photos and treat off-platform wire transfers or cryptocurrency requests as immediate red flags, ensuring they use a credit card with chargeback protection for all transactions.</p> <p>The researchers also cautioned against public QR codes at events and viewing parties, which are frequently used by cyber criminals to redirect users to <a href="https://www.computerweekly.com/news/366605874/Phishing-links-becoming-bigger-threat-than-email-attachments">credential-harvesting phishing sites</a>. On the mobile front, fans are advised to keep their devices patched, use reputable virtual private networks (VPNs) or cellular data when accessing public Wi-Fi, and disable automatic network joining.</p> <p>Finally, users should avoid sideloading Android applications and ensure any World Cup-related application is cross-checked against FIFA’s published list of official apps before downloading.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about cyber security in sports</h3> <ul class="default-list"> <li>Informa TechTarget editors discuss the <a href="https://www.techtarget.com/searchsecurity/video/Lessons-in-incident-response-from-the-Olympics-World-Cup">prevalence of cyber attacks on global sporting events</a> and how the challenges these events face are the same as those of everyday organisations.</li> <li>Oracle Red Bull Racing is <a href="https://www.computerweekly.com/news/366579832/How-Oracle-Red-Bull-Racing-guards-against-cyber-threats">tapping managed security services, conducting penetration tests and improving security awareness</a> among employees to fend off cyber threats such as phishing and ransomware.</li> <li><a href="https://www.computerweekly.com/news/366561697/Report-reveals-sorry-state-of-cyber-security-at-UK-football-clubs">UK football clubs demonstrate a critical lack of cyber resilience</a>, putting the data of fans and players at risk from myriad potential threats.</li> <li>The user names and passwords of Tokyo 2020 ticket holders and event volunteers were <a href="https://www.computerweekly.com/news/252504456/Tokyo-2020-hit-by-data-breach">reportedly compromised</a>, but a government official claims the data leak was not large.</li> </ul> </div> </div> </section> With the tournament underway across North America, Palo Alto Networks warns that temporary supplier ecosystems, vulnerable municipal infrastructure and geopolitical tensions are creating risks for enterprises and fans https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/sport-football-1-adobe.jpeg https://www.computerweekly.com/news/366644594/2026-World-Cup-billed-as-largest-entertainment-attack-surface-in-history Wed, 17 Jun 2026 02:25:00 GMT <p>Governments across the Asia-Pacific (APAC) region have been doubling down on <a href="https://www.computerweekly.com/opinion/Why-Asia-needs-its-own-model-of-digital-sovereignty">sovereign artificial intelligence (AI)</a> initiatives to protect national security, cultural values and data privacy. However, highly restrictive policies that lock out global cloud and AI providers could have an impact on national economies, according to a report by Oxford Economics.</p> <p>The <a href="https://www.oxfordeconomics.com/resource/the-eonomics-of-sovereign-ai-balancing-autonomy-innovation-and-growth-in-the-asia-pacific/">report</a>, commissioned by the AI Adoption Initiative, a global community of AI policy experts, warned that while building domestic AI ecosystems is positive, mandating full technological self-sufficiency will result in skyrocketing infrastructure costs and delays in enterprise AI adoption.</p> <p>In the most restrictive scenarios, where governments mandate a domestically owned, full AI stack, large economies like Japan and India could face additional direct costs of $149.7bn and $102.5bn, respectively, between 2025 and 2035.</p> <p>Speaking to Computer Weekly, Henry Worthington, managing director at Oxford Economics, noted that the research was designed to quantify the impact of policy decisions around sovereign AI. “This is one of the first efforts to model and put numbers around the potential trade-offs that you might be making as a government, depending on your position,” he said.</p> <p>The report categorises sovereign AI policies into five levels of restrictiveness, from control-and-choice policies, which maintain access to global cloud providers and apply <a href="https://www.computerweekly.com/feature/Auditing-classifying-and-building-a-data-sovereignty-strategy">data residency requirements</a>&nbsp;to a narrow set of highly sensitive workloads, to costly ownership-centric policies that maximise formal control and strategic autonomy.</p> <p>While the direct costs of building physical AI datacentres, procuring graphic processing units (GPUs) and training local talent are huge, Worthington noted that the actual economic damage comes from lost productivity.</p> <p>“What drives probably a larger proportion of the GDP loss, particularly in more restrictive policy scenarios, is the delay that they create in facilitating AI adoption within the enterprise sector and public sector organisations,” Worthington explained.</p> <p>“It is that adoption which will ultimately drive the long-run economic benefit that you would expect from this type of general-purpose technology. What you’re getting if you go down a highly sovereign route is almost inevitable delays to the best technology being available to resident entities,” he added.</p> <p>Under the highest restriction levels, the report estimates that AI adoption among firms could be delayed by three to five years, causing a loss in productivity. For Japan alone, this opportunity cost translates to a cumulative loss exceeding $58.2bn by 2035.</p> <p>Beyond GDP and productivity, the report pointed to an often-overlooked consequence of strict AI sovereignty: environmental toll.</p> <p>Hyperscale cloud providers achieve high energy efficiency through purpose-built infrastructure, advanced cooling and economies of scale. Forcing AI workloads into smaller, fragmented, domestic datacentres directly forfeits these efficiency benefits.</p> <p>The report projected that in restrictive scenarios, the duplication of infrastructure will lead to significantly higher carbon emissions and water consumption, a critical issue in APAC, where many economies still rely heavily on carbon-intensive power grids and face high water stress.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> What drives probably a larger proportion of the GDP loss, particularly in more restrictive policy scenarios, is the delay that they create in facilitating AI adoption within the enterprise sector and public sector organisations </figure> <figcaption> <strong>Henry Worthington, Oxford Economics</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <section class="section main-article-chapter" data-menu-title="A tale of different nations"> <h2 class="section-title"><i class="icon" data-icon="1"></i>A tale of different nations</h2> <p>The APAC region is currently fragmented when it comes to policies around AI sovereignty. Worthington pointed to Japan and Singapore as examples of nations walking the tightrope successfully.</p> <p>“Japan has taken the bull by the horns and adopted what we would view as a very sensible approach, and early anecdotal evidence shows that it is accelerating adoption within the business community,” Worthington said, referring to Japan’s light-touch approach to regulation aligned with international norms.</p> <p>He also praised Singapore for identifying areas of comparative advantage and maximising the AI opportunity through a hybrid sovereignty model that places controls where risks are deemed the highest while continuing to rely on global providers. Singapore’s green datacentre roadmap was also singled out in the report for balancing high-density compute expansion with net-zero emissions targets.</p> <p>Conversely, South Korea was highlighted as a nation experiencing the friction of strict sovereignty policies. Through its Cloud Security Assurance Program (CSAP), South Korea has historically imposed strict local requirements that effectively restricted foreign providers from public sector workloads.</p> <p>“Korea stands out as an example of a country that is still pursuing the type of policies which we identify as those creating economic costs and trade-offs,” Worthington observed, though the report noted that Korea has recently begun to open some procurement to global providers.</p> <p>While China is an obvious example of a successful, highly restrictive sovereign technology market, Worthington said the AI powerhouse was excluded from the research because its massive structural scale makes its model largely unreplicable for other APAC economies.</p> <p>When asked if the report dismisses the need for national AI development, Worthington stressed that it does not oppose domestic AI strategies that support local investment and talent development, but rather the exclusionary policies that cut nations off from the global technology frontier.</p> <p>Ultimately, the report suggested that for most APAC countries, sovereignty should not be defined by owning the entire AI stack, but by the agency to govern it. By blending global capability with local control, governments can protect their interests without sacrificing the economic boom promised by the AI revolution.</p> <p>“Good policy will have different levels of safeguards required, depending on the use case,” Worthington said. “Set objectives and have a high-quality monitoring system that enables business operators – whether they be sovereign or non-sovereign – to figure out how those can be achieved most effectively, rather than trying to be very rules-based and prescriptive.”</p> <p>This will involve the use of verifiable safeguards, such as data residency requirements, encryption with key management and operator accountability, allowing countries to configure AI systems on domestic terms while maintaining access to the AI innovation and cyber security infrastructure of global hyperscalers.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li>StarHub is building a <a href="https://www.computerweekly.com/news/366643695/StarHub-to-trial-SIM-based-IDs-for-governing-AI-agents">trust layer that will assign unique identities to AI agents</a>, allowing it to monitor and block malicious agentic activity in real time.</li> <li>Alibaba Cloud <a href="https://www.computerweekly.com/news/366643330/Alibaba-unveils-Qwen-37-Max-at-inaugural-Singapore-conference">debuts AI model capable of extended autonomous tasks</a>, alongside a major upskilling initiative backed by the Singapore government to ensure no jobless growth in the age of AI.</li> <li>As AI becomes increasingly capable, tech leaders at Singapore’s ATxSummit urge governments and industry to <a href="https://www.computerweekly.com/news/366643439/AI-safety-cannot-wait-for-a-Chernobyl-moment-experts-warn">build safety and accountability into AI systems</a> before a major disaster strikes.</li> <li>At Dell Technologies World, APAC tech leaders reveal how they are <a href="https://www.computerweekly.com/news/366643327/How-APAC-companies-are-rewiring-their-tech-for-the-AI-era">relying on hyperconverged infrastructure and digital sovereignty</a> to shield themselves from supply chain shocks.</li> </ul> </div> </div> </section> Oxford Economics report reveals that pursuing total AI self-sufficiency will lead to economic trade-offs, delayed enterprise adoption and higher carbon footprints across the Asia-Pacific region https://cdn.ttgtmedia.com/visuals/German/Ai-KI-robot-hand-globe-human-hand-PB-Studio-Photo-Adobe.jpg https://www.computerweekly.com/news/366644332/Strict-sovereign-AI-policies-could-cost-APAC-economies-billions Fri, 12 Jun 2026 02:57:00 GMT <p>An <a href="https://www.computerweekly.com/feature/Getting-started-with-agentic-AI">artificial intelligence (AI) agent</a> can’t tell you how much a server costs if it can’t access the price list. But when that price list is highly restricted corporate data, governing which agents can see it – and proving they won’t misuse it – has become one of the toughest problems in enterprise AI.</p> <p>In a recent interview with Computer Weekly, Michael Gerstenhaber, vice-president of product management for agent platform at Google Cloud, illustrated the challenge using an internal Google example involving two generations of <a href="https://www.techtarget.com/whatis/definition/tensor-processing-unit-TPU">tensor processing unit</a> (TPU) hardware, codenamed Viperfish and Ghostfish.</p> <p>Gerstenhaber noted that if a user asks an AI agent to calculate the financial conversion ratio between the two systems, the agent cannot simply guess – it requires access to a highly restricted internal rate card.</p> <p>“That rate card is very sensitive, and you have to have a certain level of privilege to see it,” he explained. “It’s only through identity, permissioning, audit and observability that I’ll ever be comfortable giving my virtual employee access to sensitive data – because that’s how we treat real employees.”</p> <p>Scaling autonomous AI agents requires looking beyond the capabilities of smarter <a href="https://www.techtarget.com/searchenterpriseai/news/366644013/Trump-AI-order-targets-frontier-model-prerelease-review">frontier models</a>. Success hinges on how well an organisation embraces <a href="https://www.computerweekly.com/news/366637674/Singapore-debuts-worlds-first-governance-framework-for-agentic-AI">AI governance</a> practices, including agent lifecycle management and data access policies, to build trust in the technology.</p> <section class="section main-article-chapter" data-menu-title="‘Safe by default’ and preventing data exfiltration"> <h2 class="section-title"><i class="icon" data-icon="1"></i>‘Safe by default’ and preventing data exfiltration</h2> <p>Gerstenhaber advised organisations to approach <a href="https://www.computerweekly.com/opinion/Generative-and-agentic-AI-in-security-What-CISOs-need-to-know">agentic AI security</a> with a philosophy that mirrors human corporate accountability: software must be safe by default.</p> <p>“An employee should have to show good judgement, and that means an agent should have to show good judgement,” he said. “If an employee tries to maliciously exfiltrate data, they should be held accountable – but if they don’t try, it should still be very, very hard to exfiltrate data.”</p> <p>To enforce this, Google Cloud has deployed an agent gateway that allows administrators to set overarching corporate policies. This <a href="https://www.techtarget.com/searchsecurity/definition/defense-in-depth">defence-in-depth</a> approach combines dedicated agent registries, skills libraries and a <a href="https://www.techtarget.com/searchenterpriseai/tip/How-the-Model-Context-Protocol-simplifies-AI-development">model context protocol (MCP)</a> registry. Together, these tools ensure that even if an AI agent builds a flawed workflow, the enterprise-wide policy will step in to block unauthorised access.</p> <p>“If you’re going to access a sensitive database, you really want to do it the same way every time, and you want that to be governed,” Gerstenhaber noted.</p> </section> <section class="section main-article-chapter" data-menu-title="Managing agent vs human identities"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Managing agent vs human identities</h2> <p>While AI agents have been described as digital workers, governing agents is completely different from people management. Blending human and agent management into a single dashboard presents unique challenges because their risk profiles are fundamentally different, Gerstenhaber said.</p> <p>“Agents are infinitely scalable, and Michael is not. Agents are not afraid of getting fired, but Michael is afraid of getting fired,” Gerstenhaber joked. “The amount of judgement you allow them to express is different, and that has to be contemplated in the permissions you give them.”</p> <p>Managing these permission levels requires strict compartmentalisation. For instance, an AI agent might be granted access to a specific secret document to fulfil a task, while simultaneously being blocked from top-secret files that its human manager is cleared to see.</p> </section> <section class="section main-article-chapter" data-menu-title="The observability dilemma"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The observability dilemma</h2> <p>Maintaining visibility into an agent’s decision-making process is another bugbear. Drawing on his previous experience at observability specialist Datadog, Gerstenhaber pointed to the use of <a href="https://www.techtarget.com/searchitoperations/definition/distributed-tracing">distributed tracing</a> to track agentic workflows. Through this, administrators can audit exactly what an agent did, which tools it selected, its permission status during a query and even the internal “thoughts” of the model.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> It’s only through identity, permissioning, audit and observability that I’ll ever be comfortable giving my virtual employee access to sensitive data – because that’s how we treat real employees </figure> <figcaption> <strong>Michael Gerstenhaber, Google Cloud</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>However, presenting that dense telemetry to business leaders requires careful design. “The difficulty is not scaring the person who’s trying to interpret it," Gerstenhaber said. A sales manager, for example, shouldn’t have to grapple with complex, technical diagrams just to understand what an AI assistant is doing.</p> <p>On the security front, Google Cloud relies on <a href="https://www.computerweekly.com/news/366614834/Inside-Google-Clouds-secure-AI-framework">Model Armor</a> to protect production deployments. Operating entirely out-of-band, Model Armor monitors live interactions between the <a href="https://www.techtarget.com/searchapparchitecture/definition/application-program-interface-API">application programming interface</a> (API) generating the pre-fill and what gets decoded during inference. Because it operates outside the reach of the engineers who built the agent, it can independently guard against prompt injections and toxicity without internal interference.</p> </section> <section class="section main-article-chapter" data-menu-title="Agents may never be decommissioned"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Agents may never be decommissioned</h2> <p>When asked how enterprises should manage the agentic lifecycle, specifically commissioning, retraining and retiring AI agents, Gerstenhaber offered a different perspective: they might never need to be retired at all.</p> <p>Because the underlying foundation model powering an agent remains immutable after release, the agent’s operational behaviour can be continuously corrected without taking the system offline.</p> <p>“You don’t even have to decommission it, really,” Gerstenhaber said. By using observability traces, human managers or automated “judges” powered by <a href="https://www.techtarget.com/whatis/definition/large-language-model-LLM">large language models</a> can flag poor interactions and feed corrections directly back into the agent’s memory.</p> <p>“It’ll improve with that kind of online learning and ‘fine-tune’ the bad behaviour out of the model during runtime. It’ll get smarter and more precise even though the model itself is trained within the same generation,” he added.</p> </section> <section class="section main-article-chapter" data-menu-title="Driving towards ‘elastic intelligence’"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Driving towards ‘elastic intelligence’</h2> <p>The ultimate goal of governed, continuously learning agents is what Gerstenhaber terms elastic intelligence. For enterprise IT, this changes how work is resourced. Instead of complex tasks being bottlenecked by human hours, AI agents allow businesses to scale their operational capacity dynamically.</p> <p>“We take something that takes time, and instead it takes space or money – but it can be done infinitely quickly for the same budget,” he said.</p> <p>To realise this vision, Google Cloud is developing advanced capabilities, such as the <a href="https://aibusiness.com/generative-ai/google-aims-enterprise-cost-efficiency-with-gemini-3-5-flash">upcoming Gemini Spark personal AI agent</a> designed to run autonomously. Rather than assigning granular, step-by-step tasks, human workers will direct Spark agents based on high-level business objectives.</p> <p>“For that to happen, you have to get comfortable giving it permissions upfront so that it can run autonomously,” Gerstenhaber said. Once that trust is established, the enterprise gains a workforce that “never gets bored, never sleeps and can do a lot of highly complex work on my behalf all the time”.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li>StarHub is building a&nbsp;<a href="https://www.computerweekly.com/news/366643695/StarHub-to-trial-SIM-based-IDs-for-governing-AI-agents">trust layer that will assign unique identities to AI agents</a>, allowing it to monitor and block malicious agentic activity in real time.</li> <li>Alibaba Cloud&nbsp;<a href="https://www.computerweekly.com/news/366643330/Alibaba-unveils-Qwen-37-Max-at-inaugural-Singapore-conference">debuts AI model capable of extended autonomous tasks</a>, alongside a major upskilling initiative backed by the Singapore government to ensure no jobless growth in the age of AI.</li> <li>As AI becomes increasingly capable, tech leaders at Singapore’s ATxSummit urge governments and industry to&nbsp;<a href="https://www.computerweekly.com/news/366643439/AI-safety-cannot-wait-for-a-Chernobyl-moment-experts-warn">build safety and accountability into AI systems</a>&nbsp;before a major disaster strikes.</li> <li>At Dell Technologies World, APAC tech leaders reveal how they are&nbsp;<a href="https://www.computerweekly.com/news/366643327/How-APAC-companies-are-rewiring-their-tech-for-the-AI-era">relying on hyperconverged infrastructure and digital sovereignty</a>&nbsp;to shield themselves from supply chain shocks.</li> </ul> </div> </div> </section> With AI agents poised to act as digital co-workers, Google Cloud’s Michael Gerstenhaber argues that IT leaders must rethink identity management, security and observability to build trust in the technology https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/data-quality-risk-governance-gunayaliyeva-adobe.jpg https://www.computerweekly.com/news/366644235/Google-Cloud-unpacks-governance-challenges-of-AI-agents Thu, 11 Jun 2026 02:26:00 GMT <p>IT infrastructure and operations (I&amp;O) leaders must urgently adopt platform-centric models and brace for significant price hikes in traditional cloud services as hyperscalers look to recoup their <a href="https://www.techtarget.com/searchcio/feature/When-AI-spending-becomes-a-liability">massive investments in artificial intelligence (AI)</a>.</p> <p>Speaking at the recent Gartner IT Infrastructure, Operations and Cloud Strategies Conference in Sydney, analysts highlighted a key tension facing today’s IT leaders: managing the hype surrounding AI while maintaining core business operations and curbing costs.</p> <p>“<a href="https://www.computerweekly.com/feature/Getting-started-with-agentic-AI">AI agents</a> are at the very peak of hype, but as always, there are new, hyped-up technologies coming all the time,” said Autumn Stanish, director analyst at Gartner. “And as always, we still have to keep the lights on and do our day jobs.”</p> <p>Paul Delory, research vice-president at Gartner, noted that while foundational practices such as <a href="https://www.computerweekly.com/feature/Top-AI-infrastructure-considerations">infrastructure automation</a> and <a href="https://www.computerweekly.com/resources/DevOps">DevOps</a> remain critical, the pressure to innovate is mounting. With cost-cutting remaining the top priority for CIOs in 2026, many are banking on AI to deliver those savings.</p> <p>However, the business demand for AI is currently outpacing I&amp;O readiness. Stanish warned that half of I&amp;O leaders view integrating AI into their current infrastructure as a top challenge, creating a risk that teams could lose relevance. To prevent a repeat of the loss of control experienced during the initial rush to the cloud, Delory argued that I&amp;O must evolve into a value-driving function capable of delivering AI agents, continuous operations and platform-centric models.</p> <p>Over the coming year, Gartner advised I&amp;O teams to form dedicated AI centres of excellence and build fully automated delivery pipelines with strict cost controls. Delory pointed out that this is highly achievable within 90 days, as the necessary tools are mostly free and open source. Practical applications for AI agents in I&amp;O today include <a href="https://www.computerweekly.com/feature/Putting-AI-to-work-in-network-management">automatically responding to infrastructure changes</a> by updating scripts and playbooks, training AI to function as quality assurance engineers, and deploying compliance agents trained on human-readable policy documents.</p> <p>Moving to this platform-centric model requires organisational changes. Stanish suggested creating a dedicated platform team – absorbing traditional server and storage engineers – led by a product owner who aligns the technology directly with user needs. This demands new performance metrics, moving the focus away from foundational baselines such as uptime and towards business outcomes such as revenue growth and customer satisfaction.</p> <p>Beyond operational structures, the conference also highlighted a crisis in technology procurement. Luke Ellery, vice-president analyst at Gartner, cited a 2024 survey which revealed that 79% of buyers regretted their technology purchases, having either failed to meet expectations or settled for lesser solutions.</p> <p>To address this, Ellery urged organisations to keep senior-level sponsors engaged throughout the buying process to ensure final purchases align with business needs, rather than allowing procurement teams to simply find a cheaper but unsuitable alternative. Buyers should focus on measurable business outcomes rather than detailed feature specifications, adopting agile and lean procurement methods that allow for iterative refinement instead of rigid <a href="https://www.techtarget.com/searchsoftwarequality/definition/waterfall-model">waterfall processes</a>.</p> <p>Furthermore, Ellery advised leaders to embrace risk tolerance thoughtfully – using data to understand risks rather than blindly avoiding them – and to build confidence in supplier negotiations through targeted training and market knowledge.</p> <p>The risks of poor investment are particularly acute in IT support. Gartner predicts that by 2027, half of all AI projects designed for the service desk will be abandoned due to unforeseen costs, risks, or a failure to achieve projected returns on investment.</p> <p>To avoid becoming part of that statistic, Gartner director analyst Joe Rogus suggested focusing on easily accessible capabilities within existing software, particularly <a href="https://www.techtarget.com/searchcustomerexperience/definition/virtual-agent">virtual support agents</a> (VSAs) that can deflect incidents away from human staff.</p> <p>AI can also heavily support human agents through proprietary knowledge discovery using <a href="https://www.computerweekly.com/feature/Understanding-RAG-architecture-and-its-fundamentals">retrieval augmented generation (RAG)</a>, converting chat logs into new knowledgebase articles, and using <a href="https://www.computerweekly.com/feature/Making-sense-of-AIs-role-in-cyber-security">machine learning for endpoint anomaly response</a>. Furthermore, AI can help to categorise and route tickets, as well as automate case summarisation, provided organisations put in the effort to clean up their existing data first.</p> <p>Rogus also warned against the growing hype around agentic AI, noting that many suppliers are merely slapping the label on basic automation tools. True agentic AI requires giving systems the autonomy to take action on their own, a step that requires high confidence and rigorous data hygiene.</p> <p>Looking further ahead to the future of cloud computing, Rogus noted that public cloud spending is expected to exceed $1tn by 2027, heavily driven by AI. However, as hyperscalers pour hundreds of billions into AI infrastructure, they are expected to recoup these costs by hiking prices on traditional cloud services.</p> <p>To prove the ongoing business value of cloud migrations, Rogus pointed to the <a href="https://www.computerweekly.com/news/366632571/Infor-doubles-down-on-APAC-with-cloud-and-AI">rise of AI-infused, industry-specific, composable solutions</a> that move away from siloed infrastructure towards a core layer supporting a data fabric and packaged business capabilities.</p> <p>This carries major implications for IT strategies leading up to 2030. On the sovereignty front, IT leaders must carefully differentiate between data, operational and technological sovereignty, balancing the trade-offs between global hyperscalers and local providers. Multicloud strategies will also require a rethink, with Gartner predicting that most enterprises will eventually perform intensive AI model activity in one cloud while leveraging it with their data in another.</p> <p>Sustainability will become a bottleneck, as <a href="https://www.computerweekly.com/ezine/Computer-Weekly/Inside-the-AI-factory-of-the-future">AI-optimised datacentre racks</a> require significantly more power than traditional servers, potentially tripling energy demand by 2030. Meanwhile, security frameworks will need to evolve from static policies to dynamic, real-time approaches, as <a href="https://www.computerweekly.com/news/366642385/Oracle-is-turning-corporate-software-over-to-AI-agents">AI agents effectively serve as digital workers</a> in the network.</p> <p>Finally, <a href="https://www.computerweekly.com/news/366641816/How-the-AI-boom-is-reshaping-tech-cost-management">cloud financial management</a> will become non-negotiable. With AI workloads largely running in containers that are currently vastly over-provisioned, Gartner warned that companies failing to optimise their compute environments face paying up to 50% more than their leaner rivals.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in Australia</h3> <ul class="default-list"> <li>Melbourne-based Heidi is building its own AI models and launching wearable hardware to automate documentation and&nbsp;<a href="https://www.computerweekly.com/news/366640992/Aussie-AI-health-tech-Heidi-aims-to-cure-clinical-burnout">reduce the administrative burden on doctors</a>.</li> <li>The Australian government has struck a major&nbsp;<a href="https://www.computerweekly.com/news/366639595/Australia-inks-five-year-deal-with-Microsoft-to-drive-AI-and-cloud-adoption">five-year volume sourcing agreement with Microsoft</a>&nbsp;to speed up adoption of AI and cloud technologies across the public sector.</li> <li>ANZ Bank has started&nbsp;<a href="https://www.computerweekly.com/news/366638802/ANZ-rolls-out-AI-agents-for-business-bankers">rolling out AI agents within its new CRM system</a>&nbsp;to help business bankers recover hours of lost productivity by automating tasks and streamlining workflows.</li> <li>Oracle has&nbsp;<a href="https://www.computerweekly.com/news/366640566/Oracle-opens-Sydney-customer-excellence-centre-to-boost-AI-adoption">opened an AI customer excellence centre in Sydney</a>&nbsp;to help its customers across Australia and Oceania adopt the technology.</li> </ul> </div> </div> Organisations risk losing control of their IT infrastructure unless they embrace platform-centric models, modernise procurement and cut through the agentic AI hype, Gartner analysts warn https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/network-infrastructure-4-fotolia.jpg https://www.computerweekly.com/news/366644158/Brace-for-cloud-price-hikes-and-AI-failures-amid-pressure-to-modernise Tue, 09 Jun 2026 21:23:00 GMT <p>Artificial intelligence (AI) is advancing at a pace few organisations anticipated. As the conversation shifts from whether to use AI, to how to scale it safely and responsibly, <a href="https://www.gartner.com/en/data-analytics/topics/data-governance">data governance</a> is now firmly in the spotlight. This is revealing gaps many Australian business leaders didn’t know existed.</p> <p>The quality and trustworthiness of data now directly determine whether AI delivers value or introduces risk. Yet data governance frameworks have often failed to keep up, leaving organisations exposed as expectations around trust and accountability accelerate.</p> <p>While this sounds like a technical challenge, data governance is rarely about technology alone. It most often fails because of people, not tools. Low data-driven maturity, difficulty demonstrating business value and limited understanding of data, analytics and AI across the business remain the biggest inhibitors.</p> <p>Gartner predicts 60% of organisations that fail to address the cultural challenges associated with <a href="https://www.gartner.com/en/data-analytics/topics/data-analytics-strategy">data and analytics</a> governance by 2027 will fail to govern AI successfully.</p> <p>In addition, many still treat culture and governance as separate priorities, resulting in process heavy approaches that struggle to engage stakeholders or inspire stewardship. This leads to declining participation, as well as increased risk exposure and diminished returns from AI investments.</p> <section class="section main-article-chapter" data-menu-title="Data governance needs leadership"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Data governance needs leadership</h2> <p>As AI raises the stakes for data governance, many organisations are discovering that their current approach is no longer sufficient.</p> <p>Too often, responsibility sits within IT even as success now depends on coordinated action across the business. Without broader authority and buy-in, technology-led teams struggle to influence senior stakeholders or drive the behavioural change that effective governance requires.</p> <p>This is where executive sponsorship becomes essential. Overcoming cultural challenges, such as disengagement, competing priorities and reluctance to change, demands authority and credibility at the highest levels.</p> <p>A senior executive sponsor, whether the CEO or a non-technology leader, brings influence across business units and can clearly articulate why data governance matters, providing both mandate and motivation for meaningful engagement. This can unlock participation across functions, resolve tensions and reinforce that governance isn’t optional but a strategic imperative.</p> </section> <section class="section main-article-chapter" data-menu-title="Making data governance work"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Making data governance work</h2> <p>Data governance often struggles due to how it’s positioned, not because the rules are wrong.</p> <p>When it’s still seen as control heavy or IT‑owned, engagement quickly drops. In an AI‑driven environment, those perceptions don’t just slow progress, they actively limit the value organisations can get from their data.</p> <p>What makes a difference is how governance connects to the business. At its core, governance is about trust in data and how it’s used.</p> <p>That trust underpins better decision making, stronger compliance, operational efficiency and AI initiatives that can scale with confidence. But this only resonates when governance is framed in terms of business outcomes, not data quality tasks or technical controls.</p> <p>It also requires a shift in how responsibility is shared. Governance isn’t something one team owns. It relies on coordinated effort across policy setting, enforcement and execution, spanning both business and technology.&nbsp;</p> <p>The goal isn’t more structure or bigger committees, but clearer alignment to priorities that matter. When people understand how governance supports their goals, rather than seeing it as additional work, participation improves and governance begins to stick.</p> </section> <section class="section main-article-chapter" data-menu-title="Focus fast and prove value"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Focus fast and prove value</h2> <p>One of the fastest ways to derail a data governance programme is trying to take on too much.</p> <p>When priorities aren’t clear, organisations often default to bottom‑up data hygiene, such as cataloguing, cleansing and documenting data in isolation from real business needs. This takes time and delivers little visible value upfront, leaving governance efforts exposed to disengagement and defunding.</p> <p>A more effective approach is to start with specific outcomes, not data management issues. Governance should be anchored to a small number of business priorities, such as regulatory risk, AI initiatives, or workflows mature enough to deliver results quickly. Even an initial view of priorities helps frame discussion, demonstrates business understanding and builds momentum. The goal isn’t perfection, but alignment.&nbsp;</p> <p>Done well, data governance becomes a process of continual improvement. By focusing on a manageable set of outcomes and embedding governance into existing business workstreams, organisations can deliver early wins, build confidence and expand over time.</p> <p>This not only accelerates value, it reinforces the cultural shift needed for governance to stick, making it part of how the business operates rather than a standalone initiative.</p> <p><i>Sally Parker is a senior director analyst at Gartner, focused on master data management, data &amp; analytics strategy, data governance and data-driven culture</i></p> </section> Data governance is critical for scaling AI safely, but the biggest hurdles are people, not technology. Here’s why moving governance out of IT and securing executive buy-in are key for AI success https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/diversity-team-meeting-people-Rawpixel-adobe.jpg https://www.computerweekly.com/opinion/Navigating-culture-to-govern-AI-successfully Fri, 29 May 2026 02:54:00 GMT <p>A year may sound like a long time in enterprise technology, but in the field of <a href="https://www.computerweekly.com/resources/Artificial-intelligence-automation-and-robotics">artificial intelligence</a> (AI), the past 12 months have completely rewritten the enterprise architecture playbook.</p> <p>Speaking to Computer Weekly on the sidelines of the Dell Technologies World conference in Las Vegas, Dell’s global chief technology officer, John Roese, noted that the maturation of agentic AI is forcing IT leaders to rethink their infrastructure, data management and operational costs.</p> <p>“We have shifted our assumption in that the use of AI is no longer a one-shot task like a chatbot,” said Roese. “It’s about handing objectives to the AI system, and that’s what agents do today.”</p> <p>As an example, he pointed to Google’s redesign of its search engine. “You give it an objective, it does some search stuff, and then it builds a whole page for you,” said Roese. “Those are all agents working to accomplish an objective.”</p> <p>Because the user experience with agentic AI is far superior – the human becomes an instructor rather than a doer – enterprises are ripping up old generative AI use cases to rebuild them as agentic workflows.</p> <section class="section main-article-chapter" data-menu-title="Busting the GPU training myth"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Busting the GPU training myth</h2> <p>While the initial AI boom had fuelled the rush to secure graphics processing units (GPUs) for model training, Roese said the infrastructure requirements of enterprises are vastly different from that of hyperscalers.</p> <p>“The myth out there is that enterprises need thousands of GPUs,” said Roese. “Our biggest workload inside of Dell only sits on 16 GPUs and supports 40,000 people. You don’t need thousands of GPUs in an enterprise, because for each workload, agent or project, you only need a handful of GPUs, sometimes half a GPU.”</p> <p>That’s because much of the enterprise AI estate is entirely focused on inference, not training. “For agents, you only need inference,” he said. “There’s no training for agents.”</p> <p>That said, the architecture needed for inference workloads is changing as well. When enterprises were building chatbots, the architecture resulted in a very light central processing unit (CPU) load. AI agents, however, use external tools, communication protocols and knowledge graphs – components that do not naturally live in the GPU.</p> <p>“When you move to agentic, it’s almost balanced,” said Roese. “The number of CPUs and GPUs are very similar, about maybe for every two GPUs you have a CPU. You don’t just build an AI infrastructure with a pile of GPUs – you build it with GPUs and traditional CPU compute.”</p> </section> <section class="section main-article-chapter" data-menu-title="Air-gapped frontier models and the edge"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Air-gapped frontier models and the edge</h2> <p>Enterprises are also benefiting from changes in how powerful AI models are being deployed. A year ago, the most capable frontier models were locked behind cloud application programming interfaces (APIs).</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Our biggest workload inside of Dell only sits on 16 GPUs and supports 40,000 people. You don’t need thousands of GPUs in an enterprise, because for each workload, agent or project, you only need a handful of GPUs, sometimes half a GPU </figure> <figcaption> <strong>John Roese, Dell Technologies</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>But with hyperscalers now enabling <a href="https://www.computerweekly.com/news/366642734/Googles-Agentic-Data-Cloud-to-power-systems-of-action">top-tier models to be run on-premise through services such as Google Distributed Cloud</a>, Roese noted that a private model can now be deployed in multiple topologies. “You can consume it in a virtual private cloud or your datacentre, and you can <a href="https://www.techtarget.com/whatis/definition/air-gapping">air-gap it from everything else</a>,” he said. “We didn’t have any of those options, except the API one, a year ago.”</p> <p>Simultaneously, AI is moving to the edge in a structured way. Roese pointed to the recent emergence of agentic frameworks such as <a href="https://www.computerweekly.com/news/366640697/Why-OpenClaw-agents-are-the-next-big-enterprise-challenge">OpenClaw</a> that run natively on devices and AI PCs. “Those have finally put some structure around running agents on devices, and that’s incredibly powerful, and not a fad,” he said.</p> </section> <section class="section main-article-chapter" data-menu-title="Re-architecting the data layer"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Re-architecting the data layer</h2> <p>Meanwhile, data strategies are evolving in tandem with agentic AI developments. Roese warned that bolting standard data storage systems onto AI compute clusters is no longer enough to meet the performance demands of AI agents.</p> <p>Instead, organisations need to build knowledge and context layers comprising vector databases, graph databases and data annotation tools. These layers cannot sit isolated and must be deeply integrated into compute.</p> <p>“One of the performance bottlenecks is you can’t get data fast enough to the GPUs to do the work,” said Roese, adding that “the GPUs you’re paying for are sitting idle, waiting for data”.</p> <p>To reduce this latency, he said <a href="https://www.computerweekly.com/news/366642734/Googles-Agentic-Data-Cloud-to-power-systems-of-action">Dell’s AI data platform</a> is now plumbed into Nvidia’s Cuda-X interfaces, effectively running data layer services directly at GPU speed.</p> </section> <section class="section main-article-chapter" data-menu-title="Mastering tokenomics and model routing"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Mastering tokenomics and model routing</h2> <p>With more model deployment options available at different pricing mechanisms, IT leaders will also have to manage the cost of AI consumption – even as the cost per token is expected to decline over time. Because “there’s no path where it becomes cheaper to do AI”, enterprises must treat AI workloads as an arbitrage game, said Roese.</p> <p>Using specification-driven development – where AI writes software based on a markdown document – as an example, he noted that if an agentic framework spawns dozens of coding tasks and blindly sends them to top-tier models, enterprises could end up with a higher bill.</p> <p>But with model routing, enterprises can ensure complex planning tasks, such as creating software specifications, are sent to expensive frontier models, while routine coding tasks are routed to smaller, on-premise open-source models where energy is the only operational cost.</p> <p>“Building a piece of software and doing spec-driven development might have four or five different economic paths to ultimately get to the best overall economic efficiency,” said Roese.</p> <p>Mastering model routing, he added, will be a competitive differentiator and helps lower the cost of product development.</p> </section> <section class="section main-article-chapter" data-menu-title="The human element"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The human element</h2> <p>Ultimately, the hardest part of operationalising agentic AI relates to the human element. Roese described the traditional human job as a “container of work” that includes a mix of hygiene, productivity, coordination and expert tasks. Agents cannot perform an entire job, but they are highly capable of executing specific types of work in that container.</p> <p>Dell itself had audited 6,400 jobs across its own business to see how AI agents would impact its workforce.</p> <p>“The first thing we realised is every single job in the company is going to change,” said Roese. “I’m taking work out of the job and removing stuff from the container. If the container is now only half full, do I need half the number of people, or do I expand that by half? Am I able to do more expert work?”</p> <p>Indeed, the impact of AI on the workplace is so profound that change management has become a key remit of IT leadership.</p> <p>“For the last four months, I’ve spent 50% of my time dealing with human dynamics,” said Roese. “AI has ceased being a technology and an ROI [return on investment] discussion. It’s now very much an organisational and human dynamic discussion. You simply can’t use these things unless you fully understand how you’re going to adapt the human population around them.”</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li>Alibaba Cloud <a href="https://www.computerweekly.com/news/366643330/Alibaba-unveils-Qwen-37-Max-at-inaugural-Singapore-conference">debuts AI model capable of extended autonomous tasks</a>, alongside a major upskilling initiative backed by the Singapore government to ensure no jobless growth in the age of AI.</li> <li>As AI becomes increasingly capable, tech leaders at Singapore’s ATxSummit urge governments and industry to <a href="https://www.computerweekly.com/news/366643439/AI-safety-cannot-wait-for-a-Chernobyl-moment-experts-warn">build safety and accountability into AI systems</a> before a major disaster strikes.</li> <li>At Dell Technologies World, APAC tech leaders reveal how they are <a href="https://www.computerweekly.com/news/366643327/How-APAC-companies-are-rewiring-their-tech-for-the-AI-era">relying on hyperconverged infrastructure and digital sovereignty</a> to shield themselves from supply chain shocks.</li> <li>Nutanix CEO <a href="https://www.computerweekly.com/news/366642824/Nutanix-CEO-maps-out-agentic-AI-strategy-targets-VMware-defectors">talks up the company’s agentic AI play</a>, the growing demand for sovereign cloud capabilities, and why decoupling storage from HCI is hastening migrations from VMware.</li> </ul> </div> </div> </section> The growing adoption of agentic AI will require IT leaders to rebalance their CPU and GPU estates, tightly integrate data layers, and redesign human workflows, according to Dell Technologies CTO John Roese https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/AI-artificial-intelligence-chip-processor-2-flyalone-adobe.jpg https://www.computerweekly.com/news/366643753/Agentic-AI-is-driving-rethink-of-enterprise-architecture-and-tokenomics Fri, 29 May 2026 02:15:00 GMT <p>Kmart has partnered with Google to shore up its digital storefront with a slew of artificial intelligence (AI)-powered tools built on the tech giant’s Gemini large language models and agentic AI platform.</p> <p>This includes an AI companion, dubbed Joy, which aims to provide a conversational interface for customers that goes beyond traditional keyword search. The assistant is powered by Google Cloud’s <a href="https://www.computerweekly.com/news/366641999/Google-launches-Gemini-Agent-Platform-eighth-generation-TPUs">Gemini Enterprise for Customer Experience</a>.</p> <p>The roll-out also includes a retail-first, virtual try-on capability that allows users to preview how outfits will look on before buying. It is accompanied by a “see it in my space” feature, using AI to digitally position home products and furniture in a customer’s actual living space.</p> <p>Bernard Wilson, Kmart Group’s chief customer officer, noted that the investment in AI is being driven by changes in consumer behaviour, adding that shoppers are increasingly looking for interactive guidance rather than static product catalogues.</p> <p>“Customers aren’t just searching anymore; they’re engaging conversationally and looking for ideas and guidance,” he said.</p> <p>“Our focus is on what customers value most: inspiration and affordable everyday products. Being focused on what you need and ensuring it’s the best product for what you require is key – especially when budgets are tight for families.”</p> <p>The Joy chatbot acts as a multimodal companion, which lets users craft <a href="https://www.techtarget.com/searchenterpriseai/definition/AI-prompt">natural language prompts</a>, such as specifying a style, budget or occasion, or upload photos directly to the chat. It then pulls inventory from across Kmart’s online range, compares products from Kmart, sister-brand Target, and various global and local brands hosted on the Kmart marketplace, before displaying options side-by-side to simplify the purchasing decision.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in Australia</h3> <ul class="default-list"> <li>Melbourne-based Heidi is building its own AI models and launching wearable hardware to automate documentation and&nbsp;<a href="https://www.computerweekly.com/news/366640992/Aussie-AI-health-tech-Heidi-aims-to-cure-clinical-burnout">reduce the administrative burden on doctors</a>.</li> <li>The Australian government has struck a major&nbsp;<a href="https://www.computerweekly.com/news/366639595/Australia-inks-five-year-deal-with-Microsoft-to-drive-AI-and-cloud-adoption">five-year volume sourcing agreement with Microsoft</a>&nbsp;to speed up adoption of AI and cloud technologies across the public sector.</li> <li>ANZ Bank has started&nbsp;<a href="https://www.computerweekly.com/news/366638802/ANZ-rolls-out-AI-agents-for-business-bankers">rolling out AI agents within its new CRM system</a>&nbsp;to help business bankers recover hours of lost productivity by automating tasks and streamlining workflows.</li> <li>Oracle has&nbsp;<a href="https://www.computerweekly.com/news/366640566/Oracle-opens-Sydney-customer-excellence-centre-to-boost-AI-adoption">opened an AI customer excellence centre in Sydney</a>&nbsp;to help its customers across Australia and Oceania adopt the technology.</li> </ul> </div> </div> <p>Paul Migliorini, vice-president of Google Cloud Australia and New Zealand, said AI technology offers an opportunity for the retail sector to redefine the customer journey.</p> <p>“Built with Gemini Enterprise for Customer Experience, these new AI-powered features and Joy’s multimodal capabilities mean Kmart customers can now shop however they prefer, with a photo or text, to easily find the right products and recommendations,” he said. “Everyday tasks – like planning an event or finding the perfect gift – feel incredibly intuitive.”</p> <p>The AI shopping tools are currently live on Kmart’s website and slated to launch on the Kmart mobile app in June 2026.</p> <p>Besides Kmart, hardware chain <a href="https://www.computerweekly.com/news/366642113/Bunnings-shows-off-AI-shopping-agent-at-Google-showcase">Bunnings has built its Buddy shopping assistant</a> using Gemini Enterprise for Customer Experience to provide customers with expert advice, helping them find what they need. Customers can tell Buddy what they want to build, and it will recommend the required tools and materials while linking to how-to videos.</p> <p>Following its progressive roll-out on the Australian website, Buddy will be launched in New Zealand later this year. Bunnings also plans to consolidate its customer service touchpoints so that it handles initial support queries.</p> <p>Additionally, it intends to integrate customer loyalty data, enabling the shopping assistant to offer hyper-personalised recommendations with customer consent, such as automatically suggesting tools from brands that a customer is already using.</p> The retailer is deploying Google Cloud’s AI capabilities to let customers preview clothes on themselves and visualise furniture in their homes as it embraces conversational commerce to win over shoppers https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/fashion-clothes-retail-shopping-teenagers-adobe.jpeg https://www.computerweekly.com/news/366643677/Kmart-taps-Google-AI-to-launch-virtual-try-ons-in-retail-first Thu, 28 May 2026 00:10:00 GMT <p>As <a href="https://www.computerweekly.com/resources/Artificial-intelligence-automation-and-robotics">artificial intelligence (AI) deployments</a> outgrow their experimental phase, Asia-Pacific organisations are now leaning on infrastructure standardisation and sovereign AI to scale globally and dodge supply chain bottlenecks.</p> <p>Speaking at a media briefing on the sidelines of Dell Technologies World 2026 in Las Vegas, executives from Dell, <a href="https://www.computerweekly.com/news/366622934/Standard-Chartered-grounds-AI-ambitions-in-data-governance">Standard Chartered</a> and South Korean tech giant <a href="https://www.computerweekly.com/news/252516460/Naver-outlines-global-AI-ambitions">Naver Cloud</a> discussed how the maturity of the AI cycle is reshaping datacentres across the Asia-Pacific (APAC) region.</p> <p>While Dell revealed that its <a href="https://www.techtarget.com/searchdatacenter/news/366643254/Dell-AI-Factory-gets-rack-scale-infrastructure-refresh">AI factory</a> customer base has surged from 3,000 to over 5,000 in the past year, beneath the rapid adoption of AI software and Nvidia-powered infrastructure lies a deeper need for resilient, highly commoditised infrastructure.</p> <p>For Standard Chartered, operating across 54 global markets required the bank to redesign its <a href="https://www.techtarget.com/searchcloudcomputing/definition/private-cloud">private cloud</a> infrastructure to achieve global scale and survive hardware shortages, according to its global head of infrastructure and operations, John Sharratt.</p> <p>The bank operates 52 country datacentres and four global datacentres as it looks to become a “super connector” for clients operating across borders.</p> <p>To manage the bank’s infrastructure footprint, Sharratt’s team eliminated all specialised hardware components in favour of a fully virtualised, <a href="https://www.computerweekly.com/resources/Hyper-converged-infrastructure">hyper-converged environment</a>, where the standard unit of scale is the server rack itself.</p> <p>The remit of “simplicity, commodity and scale” meant that all storage, networking and security had to be hyperconverged. The bank eliminated standalone <a href="https://www.computerweekly.com/feature/AI-storage-NAS-vs-SAN-vs-object-for-training-and-inference">storage area networks</a> (SANs) and physical firewall appliances, instead relying on “boring” and interchangeable commodity hardware – specifically Dell servers and switches – housed within standard racks.</p> <section class="section main-article-chapter" data-menu-title="Supply chain shortages"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Supply chain shortages</h2> <p>By avoiding highly specialised components, Standard Chartered has shielded itself from the worst of the industry’s <a href="https://www.computerweekly.com/microscope/feature/Facing-the-pressure-caused-by-supply-chain-shortages">supply chain shortages</a>.</p> <p>“If you have a very specialist design, you cannot substitute components if the NICs [network interface cards], memory or disks are not available,” said Sharratt. “By eliminating that specialisation, we can literally roll a rack off the back of a lorry and have workloads running in just 24 hours.”</p> <p>The bold move required navigating <a href="https://www.computerweekly.com/news/366627792/Forrester-urges-IT-leaders-to-dump-technical-debt">decades of technical debt</a>, which is notoriously difficult to do in the financial services sector. Sharratt said the bank set up an architectural review board run by engineers to systematically refactor, virtualise and scale every legacy application.</p> <p>“There’s always lots of legacy in a banking environment,” he said. “We have resolved all that legacy. There’s no physical server; it is all virtualised. We are sitting in an extremely comfortable position as a bank to ride out the supply chain problems.”</p> <p>Sharratt revealed that the bank’s estate in Asia, which accounts for 70% of its global infrastructure footprint, is already running on this new architecture, with the model currently being deployed in the UK. “This is not a story of what we’re going to do,” he added. “This is a story of what we’ve done.”</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Infrastructure and models only matter if they solve real-world problems. Ultimately, what we want to create is unprecedented practical value that does not exist in the world today </figure> <figcaption> <strong>Kim Yu-won, Naver Cloud</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> </section> <section class="section main-article-chapter" data-menu-title="Naver Cloud takes sovereign AI global"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Naver Cloud takes sovereign AI global</h2> <p>While Standard Chartered focuses on standardising infrastructure, Naver Cloud is leveraging its massive domestic datacentre footprint to export sovereign AI capabilities globally.</p> <p>Naver, South Korea’s leading IT portal and one of the few global search engines to successfully defend its home turf against Google, operates massive infrastructure, including the multi-megawatt <a href="https://junglim.com/en/naver-datacenter-gak-sejong/">Gak Sejong datacentre</a>. Having built its own generative AI model, <a href="https://clova.ai/en/hyperclova">HyperClova X</a>, the company is now partnering with Dell and Nvidia to deploy AI factories tailored for digital sovereignty.</p> <p>Kim Yu-won, CEO of Naver Cloud, noted that owning the full stack – from datacentres and graphics processing units to the underlying AI models – gives the company a unique advantage as governments and highly regulated industries look to protect their data.</p> <p>“In a world where sovereignty is key, we are highly flexible in giving each customer specific security and governance,” said Kim. “We want to provide customers that prioritise sovereignty with a dedicated private cloud, and on top of that, we want to integrate with AI technology.”</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li>As embodied AI moves from proof of concept into real-world pilots, industry leaders at the ATxSummit conference in Singapore warn that <a href="https://www.computerweekly.com/news/366643344/Embodied-AI-steps-out-of-the-lab-but-scaling-challenges-remain">large-scale enterprise adoption hinges on safety, cost and data governance</a>.</li> <li>Dell Technologies’ chief operating officer Jeff Clarke offers a <a href="https://www.computerweekly.com/news/366643493/Avoid-expensive-AI-agents-with-these-five-design-imperatives">blueprint for the AI-native enterprise</a>, warning that failing to integrate data and control tokenomics will result in high cloud bills and fragmented tools.</li> <li>While fully autonomous hacking bots remain a distant reality, an ESET expert warns that <a href="https://www.computerweekly.com/news/366642914/ESET-Dont-fear-the-AI-Terminator-but-prepare-for-agent-risks">AI is quietly supercharging phishing schemes</a> and creating new vulnerabilities inside organisations.</li> <li>Nutanix CEO <a href="https://www.computerweekly.com/news/366642824/Nutanix-CEO-maps-out-agentic-AI-strategy-targets-VMware-defectors">talks up the company’s agentic AI play</a>, the growing demand for sovereign cloud capabilities, and why decoupling storage from HCI is hastening migrations from VMware.</li> </ul> </div> </div> <p>Naver Cloud is already taking this strategy beyond Korea, <a href="https://www.koreaherald.com/article/10494320">partnering with Thailand’s Siam AI</a> to develop a Thai <a href="https://www.techtarget.com/whatis/definition/large-language-model-LLM">large language model</a>&nbsp;and launching a joint venture in Saudi Arabia to build digital twin capabilities.</p> <p>“Infrastructure and models only matter if they solve real-world problems,” said Kim. “Ultimately, what we want to create is unprecedented practical value that does not exist in the world today.”</p> <p>Dell Technologies’ newly appointed leader for the APAC region, Richard McLaughlin, said helping customers navigate supply chain challenges and chart a path to becoming AI-driven companies remain his top priorities.</p> <p>“The AI ecosystem is changing as it is being embraced and created by our customers in the AI economy,” said McLaughlin. “New business models, products and services are emerging in the region as agentic frameworks become more commonplace. We’re seeing the region’s enterprises accelerating, advancing, innovating and entering agent lifecycle development at pace.”</p> <p>He added that Dell is actively working with enterprise clients on “demand shaping” over the next four to five years to improve supply chain resilience, echoing Sharratt’s sentiment that disciplined infrastructure planning is critical.</p> <p>“We believe that we have the supply chain advantage in the marketplace,” said McLaughlin. “We have over 40 years of relationships with suppliers to help de-risk the supply chain for our key customers.”</p> </section> At Dell Technologies World, APAC tech leaders reveal how they are relying on hyperconverged infrastructure and digital sovereignty to shield themselves from supply chain shocks https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/Standard-Chartered-office-London-PR-hero.jpg https://www.computerweekly.com/news/366643327/How-APAC-companies-are-rewiring-their-tech-for-the-AI-era Mon, 25 May 2026 01:04:00 GMT <p>Enterprises that have been rushing to adopt <a href="https://www.computerweekly.com/feature/Getting-started-with-agentic-AI">agentic artificial intelligence (AI) systems</a> are consuming tokens at an unprecedented rate, resulting in exorbitant monthly bills from the major public cloud suppliers.</p> <p>Dell Technologies is looking to capitalise on the ensuing bill shock with new hardware and software offerings unveiled at its customer conference this week, betting that the future of enterprise AI is local, secure and shielded from variable cloud pricing.</p> <p>“What we’re starting to see with our customers is that the amount of tokens generated is increasing faster than token costs are coming down, which means that the overall bill for customers is going up very high,” said Varun Chhabra, senior vice-president of infrastructure solutions group at Dell, during a media briefing ahead of the conference.</p> <p>To illustrate the point, Jon Siegal, senior vice-president for Dell’s client solutions group, noted that a single developer within Dell recently burned through one billion tokens in 24 hours, racking up a $3,400 cloud bill in a single day.</p> <p>In response, Dell is introducing Dell Deskside Agentic AI, an on-premise sandbox for building, testing and running AI agents locally. Powered by <a href="https://www.computerweekly.com/news/366640697/Why-OpenClaw-agents-are-the-next-big-enterprise-challenge">Nvidia NemoClaw</a> and running on high-performance Dell workstations capable of supporting models from 30 billion up to a trillion parameters, the offering ensures sensitive data never leaves the corporate environment.</p> <p>Siegal noted that running agentic AI entirely on-premise with open models can reduce enterprise spend by up to 87% over a two-year horizon compared to public cloud APIs, with a break-even point in as little as three months.</p> <p>“The workstation is really becoming that free token generator for the right use cases,” Siegal explained. “Agentic AI, more than anything else, is most cost-effective when it's near the data.”</p> <section class="section main-article-chapter" data-menu-title="Bringing frontier models to the datacentre"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Bringing frontier models to the datacentre</h2> <p>Historically, the most powerful frontier models have been locked behind public cloud walls. But Dell looking to tear down those walls through a series of high-profile partnerships, bringing advanced models on-premise or into hybrid settings for data sovereignty and performance.</p> <p>Dell announced that <a href="https://www.computerweekly.com/news/366630071/Google-Cloud-brings-on-premises-Gemini-AI-to-Singapore">Google Gemini models will now run on-premise via Google Distributed Cloud</a> on Dell PowerEdge servers. Additionally, the company is collaborating with Palantir to bring its Foundry and AI platforms on-premises and teaming up with SpaceX AI to bring Grok’s advanced reasoning and multimodal capabilities to on-premise or hybrid environments for customers.</p> <p>“I cannot stress how big of a deal this is,” Chhabra said. “These are some of the world’s most powerful frontier models that have so far only been available in the cloud...giving customers more choice, flexibility on where they want to run these models, and bringing all of these models closer to their data and their enterprise workloads.”</p> <p>During the conference keynote, executives from major industrial and pharmaceutical giants also took the stage to detail how they are using on-premise AI infrastructure.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> AI is not just changing technology, it’s changing the economics of technology in favour of enterprise infrastructure. Now is the time to decide how you can most cost-effectively generate the tokens that you're going to need for the long term </figure> <figcaption> <strong>Michael Dell, Dell Technologies</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>Diogo Rau, executive vice-president and chief information and digital officer at Eli Lilly, explained how the company relies on a Dell supercomputer equipped with more than 1,000 Nvidia GPUs to simulate complex protein interactions for drug discovery and digitally inspect manufacturing lines in milliseconds. Meanwhile, Suresh Venkatarayalu, chief technology officer of Honeywell, described deploying physical AI servers directly at industrial sites to drive autonomous operations where real-time decision-making is critical.</p> <p>To support intensive AI models side-by-side with traditional workloads, Dell also announced a total hardware and software refresh for its flagship storage array: PowerStore Elite.</p> <p>Boasting up to three times the <a href="https://www.techtarget.com/searchstorage/definition/IOPS-input-output-operations-per-second">input/output operations per second (IOPS)</a>, density and throughput of previous generations, PowerStore Elite uses new E3 drives, removes the NVRAM cache to maximise usable capacity, and pushes Dell’s data reduction guarantee to an industry-leading 6:1 ratio.</p> <p>“The question isn’t just what can this platform do today? It is, will this decision still make sense a year from now? What happens when my workloads change? What happens when my costs shift?” Chhabra noted. “This is exactly why PowerStore Elite matters.”</p> <p>On the compute side, Dell unveiled the 18th-generation PowerEdge servers, touted as the broadest single-socket lineup the company has ever shipped. Delivering up to a 70% performance improvement over the previous generation and a 13:1 server consolidation ratio, the new servers will all ship with quantum-safe firmware in preparation for <a href="https://www.computerweekly.com/news/366640684/Shrinking-PQC-timeline-highlights-immediate-risk-to-data-security">2027 post-quantum cryptography mandates</a>.</p> <p>For organisations grappling with the physical deployment of AI fabrics, Dell also introduced the Dell PowerRack, where AI compute, network and storage are engineered as a scalable unit and the Dell PowerCool CDU-C7000, a compact cooling distribution unit delivering over 220 kW of liquid cooling capacity for high-density GPUs such as <a href="https://www.computerweekly.com/news/366636948/Nvidia-unveils-Vera-Rubin-architecture-to-power-AI-agents">Nvidia’s Rubin</a>.</p> <p>Meanwhile, Dell is also streamlining its security portfolio. The company introduced PowerProtect One, a unified cyber resilience platform that merges the capabilities of PowerProtect Data Manager and Data Domain into a single control plane, reducing deployment time by up to 75%.</p> <p>To help organisations improve resilience against cyber attacks, Dell also unveiled CyberDetect, an AI-powered analytics tool that deeply inspects data at the byte level to identify ransomware corruption. Boasting 99.99% accuracy, it allows IT teams to definitively know which data is clean after an attack, turning <a href="https://www.computerweekly.com/feature/Ransomware-All-the-ways-you-can-protect-storage-and-backup">ransomware recovery</a> “from uncertainty into AI-powered, evidence-based assurance”.</p> <p>As Dell brings these major updates to market, its message to enterprise IT leaders is clear: the infrastructure to support scalable, cost-predictable AI is ready today and the financial models of the past no longer apply.</p> <p>“AI is not just changing technology, it’s changing the economics of technology in favour of enterprise infrastructure,” Dell Technologies’ CEO Michael Dell said in his keynote address. “Now is the time to decide how you can most cost-effectively generate the tokens that you’re going to need for the long term.”</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li>With half of all GenAI projects failing after the proof-of-concept stage, organisations are discovering that technology alone isn’t a silver bullet. Here’s&nbsp;<a href="https://www.computerweekly.com/opinion/Why-50-of-GenAI-projects-fail-and-how-to-beat-the-odds">how to turn them into lasting competitive advantages</a>.</li> <li><a href="https://www.computerweekly.com/news/366642824/Nutanix-CEO-maps-out-agentic-AI-strategy-targets-VMware-defectors">Nutanix CEO Rajiv Ramaswami</a>&nbsp;talks up the company’s agentic AI play, the growing demand for sovereign cloud capabilities, and why decoupling storage from HCI is hastening migrations from VMware.</li> <li>At Qualtrics Experience Live in Sydney, leaders from Zip Co, Fonterra, Swyftx and Commonwealth Bank shared&nbsp;<a href="https://www.computerweekly.com/news/366642772/ANZ-enterprises-turn-to-AI-for-customer-and-employee-insights">how AI is accelerating research, breaking down data silos and turning feedback into measurable business value</a>.</li> <li>Monday.com’s&nbsp;<a href="https://www.computerweekly.com/news/366642592/Mondaycom-targets-third-wave-of-AI-with-OpenClaw-service">Globster service brings OpenClaw to consumers and businesses</a>&nbsp;in a bid to democratise access to agentic AI capabilities.</li> </ul> </div> </div> </section> With agentic AI driving up public cloud consumption, Dell Technologies is pitching local and hybrid infrastructure to shield enterprises from soaring token costs https://cdn.ttgtmedia.com/visuals/German/article/IT-infrastructure-1-adobe.jpg https://www.computerweekly.com/news/366643353/As-AI-costs-spiral-Dell-pitches-return-to-on-premise-datacentres Tue, 19 May 2026 13:00:00 GMT <p>For a while now, gamers have shouted at their screens, barking orders or venting frustrations at virtual squadmates who could not hear them. In the rare games that did incorporate voice commands, players were forced to memorise rigid menus containing specific phrases.</p> <p>But at the <a href="https://www.computerweekly.com/news/366640398/Nvidia-expands-Vera-Rubin-platform-details-Groq-integration">Nvidia GTC 2026 developer conference</a> earlier this year, French video game giant Ubisoft offered a glimpse into a future where onscreen characters can understand what you are saying – and talk back – through Teammates, an experimental prototype that replaces traditional, pre-programmed non-playable characters (NPCs) with squadmates powered by <a href="https://www.computerweekly.com/news/366612652/APAC-organisations-embrace-generative-AI">generative artificial intelligence (GenAI)</a>.&nbsp;</p> <p>Expanding on Ubisoft’s 2024 Neo NPC project, which was honoured under the France 2030 programme for advancing French innovation, Teammates places players in a first-person shooter alongside virtual soldiers who react to natural language, environmental context and the player’s personal slang.</p> <p>Tell your virtual teammate, “Find cover behind that car and wait for my order to shoot the closest enemy”, and the character will parse the command, evaluate its surroundings, and execute the manoeuvre while acknowledging the strategy.</p> <p>According to Ubisoft, achieving this level of immersion required more than just powerful <a href="https://www.computerweekly.com/feature/LLMs-explained-A-developers-guide-to-getting-started">large language models (LLMs)</a>. The project team had to rethink the inference pipeline to abstract complexity and optimise for latency.</p> <section class="section main-article-chapter" data-menu-title="Abstracting complexity"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Abstracting complexity</h2> <p>Behind the scenes, the goal is not just to build an AI-powered game but to build a foundation that thousands of artists, writers and designers can use without needing a background in <a href="https://www.computerweekly.com/resources/Artificial-intelligence-automation-and-robotics">AI and machine learning</a>.</p> <p>“Most game development teams don’t have all the specialised skills required to update complex GenAI systems,” Joel Gregoire, technical director at Ubisoft Paris, explained during the GTC presentation. “The answer to that question was to build a platform to abstract the complexity and make games with GenAI features.”</p> <p>Ubisoft’s solution functions as an agnostic middleware. Built around a C++ software development kit, the platform creates gameplay building blocks, such as NPC interactions, which are dynamically translated into prompts. Through custom engine plugins, this data feeds directly into Ubisoft’s proprietary Snowdrop and Anvil engines, translating raw language model outputs into engine-specific formats like facial animation data.</p> <p>“Think of it as an agnostic middleware for GenAI that we can easily plug into our in-house game engines,” said Xavier Manzanares, director of gameplay GenAI at Ubisoft. “It opens a whole lot of new opportunities for our teams.”</p> </section> <section class="section main-article-chapter" data-menu-title="The problem of the awkward pause"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The problem of the awkward pause</h2> <p>If the promise of conversational AI in gaming is exciting, the engineering required to make it convincing is formidably complex. For all their linguistic fluency, LLMs are computationally heavy and notoriously slow.</p> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Most game development teams don’t have all the specialised skills required to update complex GenAI systems. The answer was to build a platform to abstract the complexity and make games with GenAI features </figure> <figcaption> <strong>Joel Gregoire, Ubisoft Paris</strong> </figcaption> <i class="icon" data-icon="z"></i> </div> </blockquote> <p>In a normal conversation, a human responds in a fraction of a second. When Ubisoft began testing its early generative models, the characters took more than three seconds to process a player’s speech, decide on an action, generate a response, and synthesise the audio.</p> <p>“Creativity starts with quality,” Maxime Sazadaly, the technical lead machine learning engineer at Ubisoft Paris, told the audience at Nvidia GTC. “But in fact, there’s something almost as important as quality, and that’s latency.”</p> <p>A three-second delay in the middle of a virtual firefight could leave the player staring at a blank, unresponsive avatar. “Even if the action is the correct one, you won’t have a perception of intelligence just because it takes so long,” Sazadaly noted.</p> <p>To make the characters feel alive, Ubisoft’s engineers determined that the entire loop – from a player speaking into a microphone to a character reacting – had to occur in under two seconds, which the team set out to achieve in three ways:</p> <ol class="default-list"> <li><b>Using faster base models:</b> The team switched from slower models to more efficient ones, employing Nvidia’s Parakeet-tdt-v3 for automatic speech recognition (ASR), Gemini 2.5 Flash Lite for cloud LLM inference, and ElevenLabs Flash v2 for text-to-speech (TTS).</li> </ol> <ol start="2" class="default-list"> <li><b>Streaming everywhere:</b> Instead of waiting for an entire response to generate, Ubisoft implemented partial function parsing. The moment the LLM outputs its first actionable function, the data is pushed to the game’s behaviour tree so the NPC can start moving. Audio is similarly streamed and stitched together chunk by chunk.</li> </ol> <ol start="3" class="default-list"> <li><b>Prompt factorisation:</b> By identifying redundant aliases and data in their massive 10,000+ token perception prompts, the team reduced prompt sizes by 30%, significantly lowering the time-to-first-token (TTFT).</li> </ol> <p>As a result of these optimisations, the team reduced the response time to just 1.5 seconds.</p> <p>Gregoire’s team also established an <a href="https://www.computerweekly.com/feature/Enterprise-strategies-for-API-management">application programming interface</a> (API) gateway that lets developers perform inference in the cloud, accessing third-party models or Ubisoft-hosted models via Kubernetes and Nvidia graphics processing unit (GPU) operators – or entirely on-device to enable offline play and lower operational costs.</p> <p>Using <a href="https://developer.nvidia.com/rtx/in-game-inferencing">Nvidia In-Game Inferencing</a> and Cuda-in-Graphics integrations, the team successfully deployed Teammates locally on high-end consumer GPUs: the Nvidia RTX 4090 and RTX 5090. To stay within a typical AAA game’s rendering budget, they also used highly optimised <a href="https://www.computerweekly.com/feature/Early-days-for-small-language-models-and-AI-at-the-edge">small language models (SLMs)</a>, specifically the four-billion-parameter Qwen3-4B-Instruct-2507 model quantised to INT4 for speed and FP8 for quality, and the KaniTTS-400m model for local voice generation.</p> <p>“Current high-end hardware and optimised inference stacks now allow multi-model GenAI pipelines to run alongside game workloads,” said Gregoire. “Moving the inference on-device is the next logical step to make NPC interaction scalable.”</p> <p>The prototype also includes the Jaspar AI personal assistant that helps players navigate game menus, adjusts the interface for accessibility needs, like colour blindness, and offers tactical advice. Furthermore, the game’s AI constantly analyses the player’s behaviour, awarding dynamic achievements based on their playing style and providing a personalised debriefing at the end of each mission.</p> <p>Whether Ubisoft’s project will progress further to entirely replace the hand-crafted, cinematic moments that have defined blockbuster games remains to be seen. For now, the industry’s largest players view AI as a tool to bolster the gaming experience.</p> <p>“Creativity remains deeply human,” said Yves Guillemot, the co-founder and chief executive of Ubisoft. “AI provides tools that help bring creative visions to life in new ways; it can be a powerful enabler to create even more meaningful and immersive experiences for players.”</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li>With half of all GenAI projects failing after the proof-of-concept stage, organisations are discovering that technology alone isn’t a silver bullet. Here’s <a href="https://www.computerweekly.com/opinion/Why-50-of-GenAI-projects-fail-and-how-to-beat-the-odds">how to turn them into lasting competitive advantages</a>.</li> <li><a href="https://www.computerweekly.com/news/366642824/Nutanix-CEO-maps-out-agentic-AI-strategy-targets-VMware-defectors">Nutanix CEO Rajiv Ramaswami</a> talks up the company’s agentic AI play, the growing demand for sovereign cloud capabilities, and why decoupling storage from HCI is hastening migrations from VMware.</li> <li>At Qualtrics Experience Live in Sydney, leaders from Zip Co, Fonterra, Swyftx and Commonwealth Bank shared <a href="https://www.computerweekly.com/news/366642772/ANZ-enterprises-turn-to-AI-for-customer-and-employee-insights">how AI is accelerating research, breaking down data silos and turning feedback into measurable business value</a>.</li> <li>Monday.com’s <a href="https://www.computerweekly.com/news/366642592/Mondaycom-targets-third-wave-of-AI-with-OpenClaw-service">Globster service brings OpenClaw to consumers and businesses</a> in a bid to democratise access to agentic AI capabilities.</li> </ul> </div> </div> </section> Ubisoft executives offer a glimpse into the engineering behind its generative AI middleware, including the use of small language models, prompt optimisation and on-device processing to bring virtual teammates to life https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/gaming-gamer-gamification-training-sezer66-adobe.jpg https://www.computerweekly.com/news/366642940/In-video-games-of-the-future-your-AI-teammates-will-actually-listen Tue, 12 May 2026 05:15:00 GMT <p>When technology giants look to build the large datacentres required to power the internet and the <a href="https://www.computerweekly.com/news/366630422/AI-boom-to-push-Australian-IT-spending-past-A172bn">artificial intelligence (AI) boom</a>, their primary concern is not necessarily real estate, climate or tax incentives – it’s finding enough electricity.</p> <p>With Deloitte projecting <a href="https://www.deloitte.com/ap/en/perspectives/powering-asia-pacific-data-centre-boom.html">$800bn worth of datacentre investments in Asia-Pacific</a> by 2030, the availability of power has become the single most decisive factor in the region’s digital economy, according to Jean-Christophe Moureau, senior vice-president for solutions operations at Schneider Electric, a multinational company specialising in digital automation and energy management.</p> <p>“The criteria for selecting where to build a datacentre today is where we have energy,” he said in a recent interview with Computer Weekly. “And the expansions make it even more difficult, because Microsoft, Google and all the Big Tech companies are looking at developing datacentres but the energy is not there.”</p> <p>The scale of the challenge is unprecedented: global energy demand for datacentres, which was 460 terawatt-hours in 2022, is expected to reach over 1,000 terawatt-hours by 2026, according to an <a href="https://iea.blob.core.windows.net/assets/6b2fd954-2017-408e-bf08-952fdd62118a/Electricity2024-Analysisandforecastto2026.pdf">International Energy Agency (IEA) report</a>.</p> <p>Across the region, the power squeeze has already affected datacentre buildouts. Moureau noted that in Singapore, a major technology hub, energy limitations have forced delays for projects from companies such as Meta.</p> <p>“The Singapore government just said, ‘Look, sorry guys, we don’t have energy for you before 2028’,” he added.</p> <p>Against this backdrop, more datacentre operators are taking power generation into their own hands, investing in <a href="https://www.techtarget.com/searchdatacenter/tip/Pros-and-cons-of-on-site-power-generation-for-data-centers">on-site generators</a>, microgrid architectures and battery energy storage systems. In the US, major players are exploring dedicated energy parks and <a href="https://www.techtarget.com/searchdatacenter/tip/How-to-implement-nuclear-energy-for-data-centers">small modular reactors</a> to secure a steady power supply.</p> <section class="section main-article-chapter" data-menu-title="Greenfield builds"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Greenfield builds</h2> <p>In markets with untapped power potential, datacentre operators and hyperscalers are eyeing greenfield builds.</p> <p>In India, for example, multinational conglomerate <a href="https://www.reuters.com/world/india/tech-majors-commit-billions-dollars-india-ai-summit-2026-02-19/">Reliance Industries is investing $109.8bn to build AI datacentres</a>, while Microsoft is slated to open its fourth and largest datacentre region in the subcontinent later this year.</p> <p>The Middle East, specifically Saudi Arabia and the Gulf states, has also <a href="https://www.computerweekly.com/feature/Middle-East-datacentre-capacity-set-to-triple-by-2030">become a datacentre hotspot</a>, with developers increasingly proposing gigawatt-scale mega-campuses backed by sovereign wealth funds.</p> <p>Moureau noted that while new investors can “put a billion on the table in five minutes”, they lack a deep, technical understanding of datacentre operations, relying on engineering partners to translate capital into viable infrastructure.</p> </section> <section class="section main-article-chapter" data-menu-title="Optimising datacentre efficiency"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Optimising datacentre efficiency</h2> <p>The power crunch is driving the industry to optimise datacentre efficiency. At Schneider Electric, engineers are trying to tear down traditional silos of datacentre construction.</p> <p>Historically, cooling systems, IT infrastructure and electrical distribution were designed independently, but today, they must be deeply integrated and digitally monitored to shave off every possible kilowatt of wasted energy.</p> <p>One promising development is direct current (DC) power. While much of the world runs on alternating current (AC) power, converting AC to DC results in significant energy loss.</p> <p>Some players, such as ST Telemedia Global Data Centres (STT GDC), are <a href="https://www.computerweekly.com/news/366637683/STT-GDC-launches-HVDC-testbed-to-address-AIs-power-demands">piloting DC-powered datacentres</a> – a move Moureau believes could bring significant efficiency gains, drawing on his own background managing DC energy systems for submarines.</p> <p>On a macro level, geopolitical and supply chain hurdles loom large. The transition to green energy has been uneven, and the energy crisis sparked by geopolitical conflicts has led some countries to reopen coal-fired power plants. Moureau called the return to coal “the biggest failure of strategy we can face today”, adding that sustainability must remain the industry’s north star, relying more on solar energy, battery storage and microgrids.</p> <p>The biggest impediment to datacentre expansion, however, is human capital. The specialised nature of modern datacentres has created a severe labour shortage that threatens to slow development just as much as a lack of electricity.</p> <p>“We talk about energy and a lot of things, but we never talk about people,” said Moureau. “To find a cooling expert in the market is not easy. Piping is not mathematical; it’s heuristic. You need years of experience. Just try to find good welders.”</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about datacentres in APAC</h3> <ul class="default-list"> <li>With AI spurring gigawatt-scale datacentre builds across APAC, <a href="https://www.computerweekly.com/news/366641460/Optical-networks-to-bridge-the-AI-compute-consumption-gap">Ciena is deploying ultra-fast, energy-efficient optical networking</a> and AI-driven automation to ensure AI services can reach consumers.</li> <li>Digital Realty's CTO talks up the pace of AI silicon innovation, the growth of inferencing workloads, and why <a href="https://www.computerweekly.com/news/366641232/Digital-Realty-CTO-on-AI-tokenomics-and-datacentre-infrastructure">boasting about datacentre megawatts misses the point</a>.</li> <li>The APAC region is leading a global datacentre expansion, but surging energy demands and the risk of outages will require datacentre operators to adopt <a href="https://www.computerweekly.com/opinion/Why-proactive-asset-management-is-mission-critical-for-APAC-datacentres">proactive strategies to ensure a resilient and sustainable digital future</a>.</li> <li>At Gitex Asia 2025, industry leaders discuss how the computational demands of advanced AI models are forcing a <a href="https://www.computerweekly.com/news/366623023/How-AI-workloads-are-reshaping-datacentre-design">rethink of datacentre power, cooling and networking</a> infrastructure.</li> </ul> </div> </div> </section> A projected $800bn datacentre infrastructure investment in APAC alone is clashing with the limits of global power grids, forcing datacentre operators to rethink where and how they build the physical engines of the internet https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/datacentre-blue-fotolia.jpg https://www.computerweekly.com/news/366642996/Inside-the-global-datacentre-squeeze Mon, 11 May 2026 04:45:00 GMT <p>Australian organisations are expected to spend more than A$33.6bn on public cloud services in 2026, an increase of 17.9% from 2025, according to the latest forecast from technology research firm Gartner.</p> <p>Driven by the <a href="https://www.computerweekly.com/news/366627559/AI-adoption-grows-amid-falling-trust-in-AI-outputs">growing adoption of artificial intelligence</a> (AI), <a href="https://www.computerweekly.com/resources/Infrastructure-as-a-Service-IaaS">infrastructure as a service</a> (IaaS) is forecast to record the fastest growth among the main cloud segments, rising 24.1% to A$7.1bn. This will be closely followed by platform as a service (PaaS) at 20.9%.</p> <p>“AI-driven demand for high-performance cloud infrastructure is changing how Australian organisations are prioritising cloud spending this year,” said Adrian Wong, director-analyst at Gartner.</p> <p>“While AI compute demands are driving rapid IaaS growth, the ultimate goal for Australian organisations is business value. As the market shifts <a href="https://www.computerweekly.com/news/366627053/Enterprise-AI-adoption-moving-beyond-experimentation">from early AI experimentation to real-time inference and agentic AI</a>, organisations are relying heavily on robust PaaS environments to manage autonomous workflows and integrate them into core applications,” he added.</p> <p>Software as a service (SaaS) remains the largest spending category for Australian organisations in 2026, forecast to reach almost A$16.4bn. However, its 13.8% increase represents a slight slowdown in growth compared with 2025 as the market matures. Organisations are increasingly prioritising licence optimisation, absorbing slower seat growth and applying tighter scrutiny to their application portfolios.</p> <section class="section main-article-chapter" data-menu-title="Hyperscaler investments fuel local capacity"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Hyperscaler investments fuel local capacity</h2> <p>The continued surge in Australian cloud spending aligns with massive recent infrastructure investments from global hyperscalers eager to turn the country into a regional AI and cloud hub.</p> <p>In June 2025, Amazon Web Services (AWS) announced a record A$20bn investment to expand its datacentre infrastructure in Sydney and Melbourne between 2025 and 2029. The expansion is aimed at supporting complex AI workloads, supercomputing applications, and national security requirements.</p> <p>AWS’s investment follows <a href="https://www.computerweekly.com/news/366556772/Microsoft-to-invest-A5bn-in-Australia">Microsoft’s A$5bn commitment in late 2023</a> to expand its computing capacity by 250% and grow its hyperscale and AI infrastructure footprint to 29 datacentre locations across the country.</p> <p>As Australian organisations mature in their AI journey, more of them are gearing towards inference-optimised infrastructure as they fine-tune smaller, domain-specific models instead of relying on general-purpose large language models. “Many are turning to hybrid cloud architectures to push this processing to the edge, which lowers cloud costs while still supporting automation at scale,” Wong said.</p> <p>One of the AI applications that can benefit from inference-optimised infrastructure is agentic commerce, where virtual agents have reasoning capabilities to understand context, identify complex needs and act on behalf of the customer.</p> <p>In April 2026, Australian hardware chain <a href="https://www.computerweekly.com/news/366642113/Bunnings-shows-off-AI-shopping-agent-at-Google-showcase">Bunnings launched Buddy</a>, an AI-powered shopping assistant that provides customers with expert advice and helps them find what they need. Built with Google Cloud’s Gemini Enterprise for CX platform in just over six weeks, the tool combines Google AI and infrastructure with Bunnings’ deep product expertise to transform the e-commerce experience from “product search to project search”.</p> <p>Following its progressive roll-out on Bunnings’ Australian website, Buddy will be launched in New Zealand later this year. Bunnings also plans to consolidate its customer service touchpoints, so that Buddy handles initial support queries.</p> <p>The retailer also intends to integrate customer loyalty data, enabling the shopping assistant to offer hyper-personalised recommendations with customer consent, such as automatically suggesting tools from brands that a customer is already using.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about cloud and AI in Australia</h3> <ul class="default-list"> <li>Melbourne-based Heidi is building its own AI models and launching wearable hardware to automate documentation and <a href="https://www.computerweekly.com/news/366640992/Aussie-AI-health-tech-Heidi-aims-to-cure-clinical-burnout">reduce the administrative burden on doctors</a>.</li> <li>The Australian government has struck a major <a href="https://www.computerweekly.com/news/366639595/Australia-inks-five-year-deal-with-Microsoft-to-drive-AI-and-cloud-adoption">five-year volume sourcing agreement with Microsoft</a> to speed up adoption of AI and cloud technologies across the public sector.</li> <li>ANZ Bank has started <a href="https://www.computerweekly.com/news/366638802/ANZ-rolls-out-AI-agents-for-business-bankers">rolling out AI agents within its new CRM system</a> to help business bankers recover hours of lost productivity by automating tasks and streamlining workflows.</li> <li>Oracle has <a href="https://www.computerweekly.com/news/366640566/Oracle-opens-Sydney-customer-excellence-centre-to-boost-AI-adoption">opened an AI customer excellence centre in Sydney</a> to help its customers across Australia and Oceania adopt the technology.</li> </ul> </div> </div> </section> As hyperscalers such as AWS and Microsoft pour billions into Australian datacentres, Gartner predicts local public cloud spending will grow by 17.9% in 2026, driven by the growth in AI infrastructure demands https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/Australia-digital-flag-fotolia.jpg https://www.computerweekly.com/news/366642685/Australian-public-cloud-spending-to-surpass-A336bn-in-2026 Mon, 11 May 2026 04:41:00 GMT <p>The prevailing anxiety in boardrooms is that artificial intelligence (AI), fuelled by recent breakthrough <a href="https://www.computerweekly.com/opinion/Anthropics-Mythos-raises-the-stakes-for-security-validation">advancements in frontier models</a>, will soon unleash a wave of autonomous cyber attacks against corporate defences.</p> <p>Tony Anscombe, chief security evangelist at cyber security firm ESET, would like everyone to take a collective breath. “Some people I talk to think, ‘AI is attacking us’,” Anscombe said in a recent interview with Computer Weekly. “No, AI is not attacking you. We’re not quite a Terminator yet.”</p> <p>Instead of deploying omnipotent digital adversaries, modern cyber criminals are acting like efficiency experts, integrating AI tools to automate mundane but highly effective tasks: <a href="https://www.computerweekly.com/feature/Beyond-the-hook-How-phishing-is-evolving-in-the-world-of-AI">drafting flawless phishing emails</a>, mimicking executives in messages and automating the hunt for stolen credentials.</p> <p>The reason hackers haven’t unleashed fully autonomous AI models to attack networks, Anscombe argued, boils down to simple economics: they don’t need to.&nbsp;</p> <p>“There’s a lot of low-hanging fruit already,” he said, noting that basic internet scans continually reveal poorly secured remote-access systems and virtual private networks. “There are still a lot of organisations out there that publicly expose weaknesses. That means the cyber criminal has too much opportunity already.”</p> <p>But the seeds of next-generation, AI-driven malware are already being sown, sometimes by accident. Anscombe pointed to a recent incident involving researchers at a New York university who successfully developed a proof-of-concept malware that uses an AI prompt mechanism. Once inside a system, the malware could dynamically analyse the digital environment, rewrite its own script on the fly, and independently decide whether to steal data. The researchers, however, inadvertently published the source code to a public malware-testing database.</p> <p>“Once you put something out in the public domain, then somebody else can take it, reverse engineer it, modify it and reuse it for their own purposes,” Anscombe said. “Suddenly, you’ve got the work already done for cyber criminals.”</p> <p>While mainstream hackers are not yet using such tools, ESET researchers have begun tracing similar sophisticated tactics back to state-aligned hacking groups.</p> <section class="section main-article-chapter" data-menu-title="The view from the inside"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The view from the inside</h2> <p>For chief information security officers (CISOs), external hackers are only half the battle. Faced with intense pressure from chief executives and corporate boards, many senior leaders are doubling down on AI to stay ahead of competitors.</p> <p>The immediate risk, Anscombe noted, comes from well-meaning employees pasting sensitive corporate data or customer information into <a href="https://www.computerweekly.com/news/366555516/How-APAC-organisations-are-tapping-generative-AI">generative AI (GenAI) tools</a>, potentially running afoul of privacy laws.</p> <p>As such, Anscombe called for CISOs to establish clear governance policies on the use of tools such as OpenAI’s ChatGPT or Microsoft Copilot to prevent staff from sharing sensitive personal or corporate data with public models.</p> <p>When it comes to AI agents, Anscombe noted that because agents act independently, they could expand a company’s attack surface and inadvertently grant access to sensitive data or facilitate lateral movement by bad actors.</p> <p>To mitigate these risks, security teams must treat AI software less as traditional code and more as digital workers. “You need to make sure the agents have permissions in the same way that employees have limitations on their access rights,” Anscombe said. “You need to treat them, in some ways, like humans.”</p> <p>As AI risks multiply, the CISO’s job description is evolving too. “I think the CISO is fast becoming a business operations person, and they&nbsp;need to start understanding the business flow and the operational flow of the business to be able to help protect it,” Anscombe observed.</p> <p>On the operational level, <a href="https://www.computerweekly.com/opinion/How-AI-agents-are-driving-the-future-of-security-operations">security operations centres (SOCs) are starting to rely heavily on machine learning and AI</a> to filter massive amounts of telemetry and prevent human analysts from being overwhelmed. AI systems act like investigators, gathering evidence, highlighting anomalies and assigning probability scores, so the human analyst can make the final determination on sophisticated threats.</p> <p>However, organisations, particularly small and medium-sized enterprises (SMEs) that lack a dedicated team of specialised security analysts, can consider outsourcing to <a href="https://www.computerweekly.com/microscope/feature/How-to-sell-MDR">managed detection and response</a> (MDR) providers rather than treating security tooling as a compliance checklist, Anscombe said.</p> <p>“You can’t deploy something like <a href="https://www.computerweekly.com/news/252468781/How-EDR-is-moving-beyond-the-endpoint">EDR [endpoint detection and response]</a> and then forget it. It’s not a tick box. It needs to be managed and operated, otherwise it’s ineffective,” he added.</p> <p>Ultimately, Anscombe hopes to separate the existential dread surrounding AI from reality. He pointed to the Indian government’s use of facial recognition at a New Delhi train station, a project that successfully identified and reunited thousands of missing children with their parents in a matter of weeks. “We shouldn’t fear technology. We should make sure we use it responsibly,” he said.</p> <p>Part of that responsibility, he said, is dialling back the marketing hype that fuels public anxiety. “I saw recently an oven that claims to cook your dinner using AI,” he said, explaining that the appliance merely uses a basic moisture sensor to know when a cake is baked. “That’s a lookup table, not AI. The overuse of the word ‘AI’ doesn’t help the fear issue.”</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about cyber security in APAC</h3> <ul class="default-list"> <li>Singapore mobilised over 100 cyber defenders to neutralise a sophisticated APT actor which infiltrated Singtel, StarHub, M1 and Simba networks in the country’s <a href="https://www.computerweekly.com/news/366638973/Singapore-mounts-largest-ever-cyber-operation-to-oust-APT-actor">largest coordinated cyber incident response to date</a>.</li> <li>Japan’s Nikkei has confirmed a major data breach that potentially <a href="https://www.computerweekly.com/news/366634243/Nikkei-data-breach-exposes-personal-data-of-over-17000-staff">exposed the personal information of more than 17,000 employees</a> and business partners after hackers infiltrated its internal Slack messaging platform.</li> <li>Australian privacy commissioner warns that the <a href="https://www.computerweekly.com/news/366633983/Fewer-data-breaches-in-Australia-but-human-error-now-a-bigger-threat">human factor is a growing threat</a> as notifications caused by staff mistakes rose significantly even as total breaches declined 10% from a record high.</li> <li>Philippine bank <a href="https://www.computerweekly.com/news/366633428/BDO-Unibank-taps-Zscaler-to-secure-cloud-migration">BDO is shoring up its cyber security capabilities</a> to protect its data and systems as it moves more services to the cloud and expands its physical presence into remote areas of the archipelago.</li> </ul> </div> </div> </section> While fully autonomous hacking bots remain a distant reality, an ESET expert warns that AI is quietly supercharging phishing schemes and creating new vulnerabilities inside organisations https://cdn.ttgtmedia.com/visuals/German/article/artificial-intelligence-robot-adobe.jpg https://www.computerweekly.com/news/366642914/ESET-Dont-fear-the-AI-Terminator-but-prepare-for-agent-risks Fri, 08 May 2026 06:30:00 GMT <p>Organisations racing to implement <a href="https://www.gartner.com/en/insights/generative-ai-for-business">generative AI</a> (GenAI) find themselves caught between the pressure to innovate and the reality of what it takes to actually do so. As a result, Gartner research found at least 50% of GenAI projects were abandoned after proof of concept by the end of 2025.</p> <p>When applied well, GenAI can help organisations tackle complex challenges and build sustainable competitive advantage. When applied poorly, it becomes just another costly experiment.</p> <p>The single biggest reason GenAI fails isn’t with the technology itself - it’s how organisations approach implementation. Organisations that don’t establish specific success metrics and align GenAI with strategic objectives face the highest failure rates. GenAI must be treated as a business transformation initiative, not just a technology deployment.</p> <p>To realise meaningful results from GenAI investments, leaders must look beyond hype and address the core reasons many projects fail. Understanding these pitfalls and knowing how to avoid them can be the difference between wasted resources and lasting competitive advantage.</p> <section class="section main-article-chapter" data-menu-title="Lack of business value"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Lack of business value</h2> <p>The most fundamental reason GenAI projects fail is lack of business value. Many organisations fall into the trap of chasing flashy demos or deploying GenAI everywhere simultaneously. This approach dilutes resources across low-impact initiatives.</p> <p>Without clear prioritisation frameworks or defined success metrics, projects lack measurable business value, making them vulnerable when budgets tighten or executives demand proof of ROI.&nbsp;</p> <p>To succeed, organisations should create a rigorous AI use-case prioritisation framework that aligns with overall AI ambition and technical feasibility. It is essential to identify specific measurable outcomes, such as productivity gains, cost reductions and customer satisfaction, and track progress continuously.</p> </section> <section class="section main-article-chapter" data-menu-title="Data isn’t ready"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Data isn’t ready</h2> <p><a href="https://www.gartner.com/en/newsroom/press-releases/2026-04-16-gartner-says-organizations-with-successful-ai-initiatives-invest-up-to-four-times-more-in-data-and-analytics-foundations">Data quality</a> is the foundation of any successful GenAI initiative. Poor data affects every department, leading to unreliable outputs, failed <a href="https://www.computerweekly.com/feature/Understanding-RAG-architecture-and-its-fundamentals">retrieval augmented generation</a> (RAG) implementations and models that can't be fine-tuned effectively.</p> <p>Building an&nbsp;AI-ready data&nbsp;foundation is critical for scaling GenAI efforts. This means curating accurate, enriched and well-governed data across the enterprise, while investing in training teams on specialised data management for GenAI use cases.</p> <p>Specifically, organisations should focus on creating robust pipelines for RAG, organising and retrieving information with knowledge graphs. These all contribute to more reliable outcomes.</p> </section> <section class="section main-article-chapter" data-menu-title="Escalating total cost of ownership"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Escalating total cost of ownership</h2> <p>Rising costs kill projects even when they’re technically successful and delivering user value. What appears as negligible per-token expenses during pilots can become a total cost of ownership nightmare when multiplied across thousands of users and hundreds of use cases.</p> <p>Organisations often underestimate GenAI’s operational expenses due to limited visibility into how costs scale. Projects that appear viable in proof of concept become budget black holes in production, leading to abrupt cancellation.&nbsp;</p> <p>To avoid this outcome, <a href="https://www.computerweekly.com/news/366641816/How-the-AI-boom-is-reshaping-tech-cost-management">GenAI FinOps&nbsp;practices</a> should be adopted from day one. Educate all stakeholders – not just IT – on cost implications tied to deployment approaches, model selection and token usage, and avoid unnecessary model customisation, which can be expensive.</p> <p>It is also important to apply prompt caching strategies to reduce redundant <a href="https://www.computerweekly.com/feature/Enterprise-strategies-for-API-management">application programming interface</a> (API) calls; use model routing to route queries to appropriately sized models; and monitor costs continuously with proper allocation and visibility tools.</p> </section> <section class="section main-article-chapter" data-menu-title="Responsible AI as an afterthought"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Responsible AI as an afterthought</h2> <p>Neglecting <a href="https://www.computerweekly.com/feature/Why-responsible-AI-is-a-business-imperative">responsible AI</a> exposes organisations to regulatory violations, reputational damage, user harm and project shutdowns – risks that resonate with the C-suite and the board.</p> <p>GenAI perpetuates existing AI risks while introducing new ones like deepfakes and hallucinations. Without robust controls around safety, privacy, accountability and fairness, these risks multiply quickly.</p> <p>Responsible AI must be central from the beginning. day one. This means focusing on safety through the prevention of harmful outputs and ensuring model reliability, as well as privacy by protecting sensitive information. Accountability by establishing clear governance and ownership is also important, as well as fairness to avoid bias while ensuring equitable outcomes for all stakeholders involved.</p> <p>Equally important is implementing critical tools, like <a href="https://www.computerweekly.com/feature/Assessing-the-risk-of-AI-in-enterprise-IT">model input validation and filtering</a>; output monitoring and observability systems; compliance tracking and audit trails; and security controls for data and model access. Defining where GenAI shouldn’t be used is also an important consideration to protect against predictable disasters.</p> </section> <section class="section main-article-chapter" data-menu-title="Poor change management"> <h2 class="section-title"><i class="icon" data-icon="1"></i>Poor change management</h2> <p>Without change management, even technically excellent GenAI tools see minimal adoption. Usage drops over time. Employees feel threatened rather than empowered. The organisation captures a fraction of potential value, while technical teams wonder why their capable solution sits unused.&nbsp;</p> <p>Change management must be treated as a first-class requirement, not an afterthought. Leaders need to build empathy maps that reveal how GenAI impacts roles throughout their organisation, so they can focus on amplifying human capabilities instead of threatening job security.</p> <p>To make it easier for employees to adopt GenAI, build it into existing workflows if possible rather than requiring them to use new tools and processes. Also, involve them in the pilot to ensure the user experience is acceptable to them and make changes based on feedback. This will increase the chances of them actually using the technology in the long-term.</p> <p><i>Arun Chandrasekaran is a distinguished vice-president analyst at Gartner, specialising in AI. </i></p> </section> With half of all generative AI projects failing after the proof-of-concept stage, organisations are discovering that technology alone isn't a silver bullet. Here’s how to turn them into lasting competitive advantages https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/generative-AI-Chat-GPT-Timon-adobe.jpg https://www.computerweekly.com/opinion/Why-50-of-GenAI-projects-fail-and-how-to-beat-the-odds Fri, 08 May 2026 02:43:00 GMT <p>The days of the glorified artificial intelligence (AI) chatbot are over, with more enterprises <a href="https://www.computerweekly.com/feature/Moving-agentic-AI-from-innovation-theatre-to-enterprise-production">starting to embrace the era of autonomous agents</a> rummaging through corporate data.</p> <p>At the recent Google Cloud Next 2026 conference in Las Vegas, Google took the wraps off its Agentic Data Cloud, a rebranding and architectural effort to support the transition from passive systems of intelligence to autonomous systems of action.</p> <p>“Everyone’s talking about systems of intelligence, but frankly, they’re still ingesting all the data, looking at the past or maybe trying to predict the future,” said Andi Gutmans, general manager and vice-president for Data Cloud at Google.</p> <p>“We’re moving from human scale to agent scale, both in the number of agents we’re going to have and the workloads we have to manage,” he told Computer Weekly, referring to the exponential growth in compute requirements of agentic AI.</p> <p>To address the workload, cost and governance issues associated with agentic AI, Google unveiled around 80 new product updates, focusing on metadata management, cross-cloud interoperability and distributed database capabilities.</p> <p>A primary concern for enterprise data teams deploying AI agents is ensuring models access the right data while respecting access controls. Gutmans noted that simply knowing where data resides no longer suffices – agents need semantic context to avoid hallucinations and errors.</p> <p>“It’s typical for customers to have 500 customer tables, but which table do you want to look at?” Gutmans said. In response, Google launched the <a href="https://www.techtarget.com/searchdatamanagement/news/366641929/Google-unveils-data-cloud-purpose-built-for-agentic-AI">Knowledge Catalog</a>, which builds on its previous Dataplex governance capabilities and aggregates metadata from within Google Cloud, external cloud applications and third-party catalogues. It then enriches the metadata to map relationships across structured and unstructured data.</p> <p>Gutmans described Knowledge Catalog as a “flywheel” built directly on top of existing access controls, ensuring agents cannot surface or act on data they are not authorised to view.</p> <p>Recognising the realities of heterogeneous IT environments, Google also introduced the Cross-Cloud Lakehouse. The offering allows enterprises to run Google’s <a href="https://www.computerweekly.com/feature/Big-data-and-Google-BigQuery-improve-cancer-drug-development-by-detecting-bacteria">BigQuery</a> and AI capabilities against data residing in Amazon Web Services and Microsoft Azure, while also providing zero-copy integrations with enterprise systems like SAP and Workday.</p> <p>Gutmans was keen to split hairs on the terminology: “It’s very important to note that the term ‘cross-cloud’ is distinct from ‘multicloud’; those who refer to multicloud are just talking about multiple single-cloud environments.”</p> <p>Another significant update was Spanner Omni. Historically, Google Cloud Spanner, a globally distributed relational database, was tethered to Google’s infrastructure, including storage, as well as GPS receivers and atomic clocks to ensure transactional consistency.</p> <p>Driven by enterprise demand for disconnected edge and on-premises deployments via Google Distributed Cloud, Google has engineered Spanner to run independently.</p> <p>“Three to four years ago, no one, including us, believed we could disconnect Spanner from Google Cloud,” Gutmans said. Now, Spanner Omni offers vector processing, search and graph capabilities for disconnected environments, which Gutmans noted is key for highly regulated use cases like on-premise fraud detection.</p> <section class="section main-article-chapter" data-menu-title="The IT pro as orchestrator"> <h2 class="section-title"><i class="icon" data-icon="1"></i>The IT pro as orchestrator</h2> <p>As part of the Agentic Data Cloud roll-out, Google released a Data Agent Kit, featuring support for <a href="https://www.techtarget.com/searchenterpriseai/news/366624572/Anthropic-intros-next-generation-of-Claude-AI-models">Claude Code</a>, Gemini CLI, Codex, and VS Code extensions. The goal is to provide developers and data engineers with the model context protocol (MCP) and tools necessary to build their own agents.</p> <p>According to Gutmans, the move towards agentic AI means that rather than writing manual pipelines or Python scripts, practitioners will engage in “intent-driven development”, which allows them to focus on defining their goals and desired outcomes while the agents handle the technical implementation.</p> <p>“Every practitioner is now becoming an orchestrator of agents,” Gutmans said. “That will be true whether you’re a business user, data scientist, data analyst, data engineer or developer. We are really trying to...put this persona in a position where they have a team of agents, and they’re orchestrating the agents.”</p> <p>Serving in a dual role where he provides data infrastructure for Alphabet properties like Search, YouTube and Gmail, Gutmans noted that Google’s own <a href="https://www.techtarget.com/searchitoperations/definition/site-reliability-engineering-SRE">site reliability engineering</a> (SRE) teams are already deploying AI agents.</p> <p>“One example is an agent that looks at support tickets. If it sees that there’s more than one support ticket in a given period of time that looks very similar, it will page us and say, ‘Hey, maybe something bigger is happening here’,” Gutmans said.</p> <p>With the slew of enhancements in Agentic Data Cloud, Gutmans didn’t shy away from throwing a bit of shade at rivals, claiming Google’s ownership of the full stack – from custom tensor processors and BigQuery to DeepMind’s Gemini models – leaves rivals in the dust.</p> <p>“If you think about other hyperscalers like Azure, they don’t have the model, so they end up having to connect to some other environment,” he claimed. As for pure-play data platform suppliers like Databricks? “They neither have the infrastructure nor the model, so they’re basically kind of assembling all this stuff.”</p> <p>Moutusi Sau, managing vice-president at Gartner, said Google’s Agentic Data Cloud reframes the hyperscaler’s data platform as a semantic and orchestration layer for agents, addressing challenges such as agent failures, which are often caused by poor data context, inconsistent semantics, and fragile integration.</p> <p>Capabilities such as zero-copy and cross-cloud access reduce data gravity and duplication, but they also increase dependence on semantic accuracy, metadata governance and performance consistency, she added.</p> <p>Ultimately, Sau called for enterprises to invest as much in semantic ownership, stewardship and validation as they did in tooling. “Without disciplined governance, enterprises risk scaling ambiguity and mistrust faster than agents scale productivity,” she said.</p> <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI in APAC</h3> <ul class="default-list"> <li>Agoda, a digital travel platform, has set its sights on <a href="https://www.computerweekly.com/news/366640804/Agoda-scales-AI-strategy-opens-new-APAC-tech-hub">becoming an AI-powered travel companion</a> as it changes how it builds software and moves its tech workforce into a new facility in Bangkok.</li> <li>Singtel and Nvidia have teamed up on a multimillion-dollar facility to <a href="https://www.computerweekly.com/news/366639492/Singtel-Nvidia-to-help-scale-enterprise-AI-deployments">help organisations scale enterprise AI deployments</a>, tackle extreme datacentre power densities, and prepare for the era of embodied AI.</li> <li>The Australian government has struck a <a href="https://www.computerweekly.com/news/366639595/Australia-inks-five-year-deal-with-Microsoft-to-drive-AI-and-cloud-adoption">five-year volume sourcing agreement with Microsoft</a> to speed up adoption of AI and cloud technologies across the public sector.</li> <li>Alibaba Group has unveiled <a href="https://www.computerweekly.com/news/366640461/Alibaba-joins-AI-agent-race-with-Wukong-launch">Wukong</a>, an AI-native enterprise platform that brings advanced agentic AI capabilities directly into business workflows.</li> </ul> </div> </div> </section> As enterprises move from reactive analytics to AI agents, Google Cloud’s data chief details new metadata, cross-cloud and database tools to help them govern and scale AI agents https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/abstract-tech-AI-data-spainter-vfx-adobe.jpg https://www.computerweekly.com/news/366642734/Googles-Agentic-Data-Cloud-to-power-systems-of-action Tue, 05 May 2026 03:36:00 GMT 60 editor@computerweekly.com