Last year, I was sitting in a meeting with the leadership team of an Australian business that had one question on everyone’s mind.

"What does AI mean for our people?"

It didn’t take long before the conversation went where almost every AI conversation seems to go.

• Will jobs disappear?
• Should we slow down hiring?
• Which roles are most at risk?

They’re fair questions. We’ve all been hearing them for the past couple of years.

But as we started looking at how their teams actually worked, something interesting happened.

• Nobody wanted to replace their marketing manager.
• Nobody wanted fewer project managers.
• Nobody was planning to remove their sales team.

Instead, we found dozens of small tasks eating away at people’s time every single week.

• Meeting notes.
• Research.
• Proposal writing.
• Reporting.
• Spreadsheet clean-up.
• Summarising documents.
• Following up emails.

One by one, AI started taking care of those tasks.

• The people stayed.
• The work changed.
• That meeting has stayed with me ever since.

Because I think we’ve been asking the wrong question.

For the past two years, we’ve asked whether AI will replace workers.

But after working with organisations introducing AI into their operations, I’ve become convinced the bigger story is something else entirely.

The workplace itself is changing.

Not overnight.

Not in one dramatic moment.

Quietly.

One workflow at a time.

One process at a time.

One team at a time.

That difference matters.

Because when you stop following the headlines and start listening to the people actually building this technology, a different picture begins to emerge.

Many of the world’s leading AI companies aren’t talking about replacing people.

They’re talking about redesigning work.

That’s a very different conversation.

Now, let’s be realistic.

AI is creating disruption.

• Some jobs will change.
• Some tasks will disappear.
• Some roles will become smaller.

Even Amazon CEO Andy Jassy has openly said AI will reduce the need for some types of work while creating demand for others.

Ignoring that would be naive.

But assuming every job simply disappears isn’t supported by the evidence either.

The reality is far more interesting.

Sponsored Ad

The shift has already started

Think back to the technologies that changed the way we work.

• Email.
• Cloud software.
• Smartphones.
• Video calls.

None of them replaced entire professions overnight.

They changed how work happened first.

AI feels different because it’s beginning to support knowledge work itself.

Today, software can help analyse information, draft reports, summarise meetings, write code, brainstorm ideas and answer complex questions in seconds.

That’s why this moment feels different.

It isn’t just helping us work faster.

It’s changing how work gets done.

Microsoft’s latest Work Trend Index gives us one of the clearest pictures of what’s happening.

Nearly half of Microsoft 365 Copilot interactions are focused on analysing information, solving problems and supporting creative thinking.

More importantly, 66% of AI users said AI allows them to spend more time on higher-value work.

Another 58% said they’re producing work they couldn’t have created just a year earlier.

I find that fascinating.

Because those numbers aren’t telling us people have become less important.

They’re telling us people’s time is being spent differently.

And perhaps the most surprising finding wasn’t about productivity.

It was about judgement.

When Microsoft asked employees which human skills become more important as AI takes on more work, quality control came first.

Critical thinking came second.

In other words, AI might generate the first draft.

You’re still responsible for deciding whether it’s any good.

That changes what being a knowledge worker actually means.

AI has moved beyond experimentation

Not that long ago, AI was something businesses experimented with.

• Someone in marketing tested it.
• A developer tried a coding assistant.
• A few people used it to summarise meetings.

Most organisations were still watching from the sidelines.

We’re past that point now.

McKinsey found employee AI adoption grew from around 30% in 2023 to 76% by 2025. AI is quickly becoming part of everyday work rather than an interesting side project.

That’s a significant shift.

The conversation is no longer about whether AI works.

• It’s about where it fits.
• How should teams use it?
• Which tasks should people still own?
• What becomes more valuable once AI is part of the workflow?

Those are much harder questions.

They’re also much more useful.

BCG estimates that around half of jobs in the United States, roughly 50% to 55%, will be reshaped by AI over the next two to three years.

Not eliminated.

Reshaped.

That single word is worth paying attention to.

Because if half the tasks inside your role change, is it really the same job anymore?

Maybe.

Maybe not.

Either way, it’s becoming a different kind of work.

And that’s exactly what I’m hearing from business leaders.

Not that people are disappearing.

But that the way we work is being rewritten.

In the next section, let’s look at what the CEOs behind Microsoft, Google, OpenAI, NVIDIA, Amazon, IBM and Salesforce are actually saying.

Because once you hear them side by side, the conversation becomes far more nuanced than the headlines would have you believe.

So what are the people building AI actually saying?

Not the headlines.

Not the social media clips.

The full conversations.

Because once you start reading interviews, company blogs and earnings calls, a different picture emerges.

The story is less about replacing people and more about changing the way people work.

That doesn’t mean every CEO agrees.

They don’t.

But there are some interesting patterns.

Take Microsoft CEO Satya Nadella.

Instead of treating AI as another way to cut costs, he’s encouraged leaders to think about how they can “reorganise the job” to better use people’s strengths. He also talks about creating workplaces built around continuous learning, where people and AI work together instead of competing.

I think that’s an important mindset.

Too many businesses start with this question:

• “What can AI replace?”

The better question is:

• “What could our people achieve if AI handled the repetitive parts of their day?”

Those are two very different conversations.

Google CEO Sundar Pichai is sending a similar message.

He’s been clear that AI will automate some work, change other roles and create entirely new careers. At the same time, Google has invested heavily in AI education, saying it has already trained more than 100 million people globally in AI skills.

That tells me Google isn’t expecting people to sit on the sidelines.

It’s expecting them to learn alongside the technology.

Then there’s NVIDIA CEO Jensen Huang.

His perspective is probably the most practical.

He said it’s unlikely that most people will lose their jobs to AI.

Instead, many will lose opportunities to someone who knows how to use AI better.

Whether you agree with that or not, it’s hard to ignore the message.

AI is becoming another professional skill.

Not unlike spreadsheets.

Or email.

Or cloud software.

The people who learn it early usually gain an advantage.

Something else caught my attention while researching this topic.

Some of the strongest warnings about AI have become more measured over time.

OpenAI CEO Sam Altman recently admitted he was “pretty wrong” about how quickly entry-level white-collar jobs would disappear. The impact simply hasn’t happened as fast as he expected.

Anthropic CEO Dario Amodei has also shifted his thinking.

Rather than focusing only on automation, he now talks about how AI can expand the work people do by taking care of routine tasks while increasing overall productivity.

That doesn’t mean the risks have disappeared.

Far from it.

Amazon CEO Andy Jassy has been very open about the fact that AI will reduce the need for some kinds of work while creating demand for others.

It’s one of the most balanced perspectives I’ve seen.

He isn’t pretending every job stays the same.

He’s also not suggesting every job disappears.

He’s acknowledging something every major technology shift has done.

Some work goes away.

New work replaces it.

Sponsored Ad

Then there are the companies making decisions that don’t quite fit the public narrative.

IBM is one of them.

If you believed every headline, you’d expect large technology companies to stop hiring graduates.

IBM did the opposite.

The company announced plans to significantly increase entry-level hiring while redesigning those roles around AI. Junior engineers are spending less time on routine coding and more time working with customers, solving problems and building the skills AI can’t easily replicate.

Salesforce is another interesting example.

While many predicted AI would reduce graduate opportunities, Salesforce announced plans to hire 1,000 graduates and interns to help build its AI products.

That decision says a lot.

Technology still needs talented people.

It just needs them doing different kinds of work.

Even Meta’s story is more nuanced than most headlines suggest.

After restructuring parts of the business, Mark Zuckerberg spoke about moving employees into AI-focused roles and creating opportunities for people to transition as the company evolved, rather than treating workforce change as a one-way process.

When you step back, a pattern starts to emerge.

• No serious business leader is claiming AI won’t change work.
• They’re also not all predicting a future without people.
• Most are describing something in between.
• A workplace where routine tasks increasingly move to AI.
• Where human judgement becomes more valuable.
• Where roles evolve instead of standing still.

That’s a much more believable future.

And if that’s the direction we’re heading, then the next question becomes even more important.

It’s not just what CEOs are saying.

It’s what their organisations are actually doing.

Because that’s where the real lessons begin.

There’s an old saying in business.

Don’t listen to what companies say.

Watch what they do.

I think that’s especially true with AI.

It’s easy for executives to talk about transformation.

It’s much harder to redesign the way thousands of people work.

But that’s exactly what’s starting to happen.

And it’s happening much more quietly than most people realise.

If you walk into most businesses today, you probably won’t see robots replacing entire teams.

You’ll see something much less dramatic.

People are spending less time on repetitive work.

• They’re writing fewer first drafts.
• They’re searching for less information.
• They’re summarising fewer meetings.

The work hasn’t disappeared.

It’s been redistributed.

That’s a subtle difference.

But it’s a huge one.

Because once AI starts taking care of routine tasks, every role begins to change.

Not all at once.

Piece by piece.

That’s why I don’t think AI is simply another software upgrade.

It’s changing the structure of work itself.

The companies seeing results aren’t just buying AI

One thing I’ve noticed while working with businesses is this.

Buying AI is easy.

Changing the way people work is hard.

Almost every organisation can subscribe to an AI tool.

Very few redesign their workflows around it.

That’s where the real opportunity is.

The research backs this up.

Deloitte found that organisations making the most progress aren’t simply adding AI to existing processes. They’re redesigning workflows, introducing new responsibilities and combining human judgement with AI capabilities. AI becomes part of how work is organised, not just another tool sitting on someone’s desktop.

BCG reached a similar conclusion.

Companies that redesigned their processes around AI were significantly more likely to report measurable business improvements, stronger productivity gains and higher employee satisfaction than organisations that simply layered AI onto existing ways of working.

That stood out to me.

Because it suggests the competitive advantage isn’t AI itself.

It’s how well an organisation adapts around it.

IBM is taking a different path

Let’s look at a real example.

A lot of people assumed AI would reduce opportunities for junior software engineers.

It’s an understandable assumption.

If AI can generate code, surely companies would need fewer graduates.

IBM reached a different conclusion.

Instead of reducing graduate hiring, the company announced plans to triple its entry-level hiring while redesigning those roles for the AI era. Junior engineers now spend less time writing routine code and more time working with customers, understanding business problems and applying judgement alongside AI.

I really like that approach.

Not because it’s guaranteed to become the industry standard.

But because it starts with a different question.

Instead of asking:

• “How do we replace junior engineers?”

IBM appears to be asking:

• “How do we make junior engineers more valuable?”

That’s a completely different mindset.

And it reflects something I’ve seen with clients as well.

The biggest gains rarely come from replacing people.

They come from removing the work that stops people doing their best work.

Sponsored Ad

Salesforce tells a similar story

Salesforce could have used AI as a reason to slow graduate recruitment.

Instead, it announced plans to hire 1,000 graduates and interns to help build its next generation of AI products.

That decision says a lot.

The technology may be changing.

But talented people are still the competitive advantage.

What’s changing is where they spend their time.

• Less repetition.
• More problem solving.
• More customer conversations.
• More creativity.
• More judgement.

Those aren’t new skills.

They’re becoming more valuable because AI is taking care of more of the routine work.

Amazon reminds us the picture isn’t all positive

It’s also important to avoid swinging too far in the other direction.

Not every company is telling the same story.

Amazon CEO Andy Jassy has been upfront that AI will reduce the need for some types of work while creating demand for others as the company gains efficiency through generative AI.

That’s an important reminder.

• Some roles will become smaller.
• Some teams will change.
• Some jobs will disappear.

Ignoring that reality doesn’t help anyone.

But neither does assuming that’s the whole story.

Because every time one task becomes automated, another usually appears.

• Someone has to implement these systems.
• Someone has to govern them.
• Someone has to redesign business processes.
• Someone has to help teams adopt new ways of working.

The work doesn’t vanish.

It moves.

And that’s the thread connecting almost every company in this research.

The organisations making the biggest progress aren’t thinking about AI as a replacement for people.

They’re thinking about AI as a catalyst to redesign how work gets done.

That’s a much bigger shift.

And for individual professionals, it raises an even more important question.

If the work around us is changing, what skills become more valuable next?

So where does that leave us?

If the workplace is changing, how do we stay relevant?

It’s probably the question I’m asked more than any other.

People assume the answer is learning better prompts or using more AI tools.

I don’t think that’s the answer.

At least, not on its own.

The more organisations I work with, the more I notice something interesting.

The businesses getting the biggest return from AI aren’t necessarily the ones with the most advanced technology.

• They’re the ones with people who know how to ask better questions.
• Who understand their customers.
• Who can make good decisions.
• Who know when to trust AI.
• And just as importantly, when not to.

The skills that matter are changing

For a long time, career growth was closely tied to knowledge.

The more information you had, the more valuable you became.

Today, information is easier to access than ever.

• AI can produce a first draft.
• Summarise a report.
• Explain a complex topic.
• Generate ideas.
• Write code.

All within seconds.

That doesn’t make expertise irrelevant.

It changes where expertise creates value.

Microsoft’s latest Work Trend Index found that as AI becomes more capable, workers see quality control of AI output and critical thinking becoming even more important. Instead of producing every piece of work from scratch, many professionals are spending more time reviewing, refining and validating AI-generated outputs.

That’s a subtle shift.

But I think it’s one of the biggest changes happening in knowledge work.

Your value increasingly comes from your judgement.

Not simply your ability to produce information.

Human skills aren’t becoming less important

One of the biggest misconceptions about AI is that it reduces the importance of human skills.

The research points in the opposite direction.

PwC found that the skills required in AI-exposed jobs are changing more than twice as fast as those in less exposed roles.

New tasks emerging in AI-enabled work rely much more heavily on judgement, empathy and creativity.

Even junior positions increasingly require capabilities that were traditionally expected from more experienced professionals.

That’s encouraging.

Because it suggests the future isn’t about competing with AI.

It’s about doubling down on the things AI struggles to do well.

• Building relationships.
• Leading teams.
• Understanding context.
• Making decisions when there isn’t a perfect answer.
• Communicating with empathy.

Those have always been valuable skills.

AI is making them even more valuable.

Technology alone isn’t enough

I’ve worked with organisations at different stages of AI adoption.

Some are only just getting started.

Others are already embedding AI into everyday operations.

One pattern keeps appearing.

The biggest challenge is rarely the technology.

It’s changing the way people work.

Buying software is straightforward.

Helping a business rethink workflows, decision-making and collaboration is much harder.

That’s why simply rolling out AI tools doesn’t automatically create results.

The organisations seeing the biggest gains redesign the work around the technology, not the other way around.

Sponsored Ad

Productivity is only part of the story

Much of the public conversation focuses on efficiency.

• How many hours AI saves.
• How much money it cuts.

Those things matter.

But they’re only part of the picture.

PwC found that organisations seeing the strongest productivity growth weren’t simply reducing costs.

They were creating new value, increasing wages and continuing to invest in people.

Deloitte also reports that organisations redesigning work around AI are seeing stronger productivity, better decision-making and improved business outcomes.

BCG found similar results, with organisations redesigning workflows around AI reporting stronger performance and higher employee satisfaction than those simply adding AI to existing processes.

To me, that’s the real opportunity.

AI shouldn’t just make us faster.

• It should help us do better work.
• Better decisions.
• Better customer experiences.
• Better products.
• Better businesses.

If we’re only measuring time saved, we’re probably missing the biggest benefit.

One final thought

I don’t think history will remember this period as the moment AI replaced people.

I think it’ll remember it as the moment work itself changed.

That’s a very different story.

The companies that succeed won’t necessarily be the ones with the biggest AI budgets.

They’ll be the ones that learn fastest.

The professionals who succeed won’t necessarily be the ones who know the most prompts.

They’ll be the ones who combine AI with sound judgement, curiosity, communication and a genuine understanding of the people they serve.

That’s much harder to automate.

And that’s exactly why it becomes more valuable.

So perhaps we’ve been asking the wrong question all along.

Instead of asking:

• “Will AI replace my job?”

Maybe we should be asking:

• “How can I create more value in a workplace where AI is part of every team?”

I’d love to hear your thoughts.

Has AI changed the work you do every day, or has it changed the way you think about the work that matters most?

Google Chrome has always been more than just a browser.

For most people, it is the front door to the internet.

You search from it. You browse from it. You shop from it. You log into business tools from it. You compare services from it. You read reviews from it. You ask questions from it.

So when Chrome starts becoming more AI-powered, it is not just a small product update.

It is a signal.

A signal that the way people search, browse, compare, and make buying decisions is changing very quickly.

Recently, an article by That Privacy Guy raised a major concern: Google Chrome may be silently downloading a large on-device AI model file, around 4GB, onto supported devices without clear user consent. The file is linked to Gemini Nano, Google’s on-device AI model, and is reportedly used for certain Chrome AI features. The article also claims that if users delete the model file, Chrome may download it again unless certain AI-related settings or flags are disabled.

Now, this story can be looked at from many angles.

Privacy. Consent. Storage. Bandwidth. Environmental impact. User trust. AI ethics. Corporate transparency.

But there is another angle that every business owner, marketer, and SEO professional should be paying attention to:

This is another sign that AI is moving from being a tool we visit, to something built directly into the platforms we already use every day.

And that will change SEO.

Not overnight. But slowly, then suddenly.

What Actually Happened?

According to the article, Chrome has been installing an on-device AI model called Gemini Nano into a folder named OptGuideOnDeviceModel. The model file is called weights.bin and is around 4GB in size.

The issue is not just that the file is large.

The bigger issue is that users may not clearly know:

• That the file was downloaded
• What it is used for
• Which Chrome AI features rely on it
• Whether it can be permanently removed
• Whether it will be downloaded again
• Whether their device is being used as part of Google’s broader AI rollout

The article argues that this creates serious concerns around consent, privacy, transparency, and even environmental impact because of the scale of Chrome’s user base. If hundreds of millions of users receive a 4GB file, the total bandwidth and energy usage becomes massive.

This is where the story gets bigger than just one file.

It shows us where browsers are heading.

Your browser is no longer just a tool that loads websites.

It is becoming an AI-powered assistant, search interface, writing tool, content summariser, shopping helper, scam detector, and decision-making layer.

That matters a lot for SEO.

Why This Matters for SEO

For years, SEO has been built around a fairly simple journey.

Someone searches on Google. They see a list of blue links. They click a website. They read the page. They contact the business or buy something.

That journey is changing.

Now, people might ask a question and get an AI-generated answer before they ever click a website.

They might use AI Mode. They might see an AI Overview. They might use ChatGPT, Gemini, Perplexity, Claude, or Copilot. They might ask their browser to summarise a page. They might rely on AI to compare businesses for them.

So the new search journey looks more like this:

Someone asks a question. AI scans, summarises, compares, and filters information. AI gives the user a direct answer. The user may or may not click through to a website.

That is the big shift.

SEO is no longer just about ranking.

It is about being understood, trusted, cited, summarised, and recommended by AI systems.

The Old SEO Game Is Getting Weaker

A lot of old SEO strategies were built around volume.

Publish more blogs. Target more keywords. Create more suburb pages. Add more service pages. Use the keyword in the heading. Add 800 words. Repeat the location. Wait for rankings.

That worked for a long time.

But AI search is going to make weak content less useful.

Why?

Because AI does not need to show ten average pages if it can summarise the best answer directly.

If your content is generic, thin, copied, vague, or written only for keywords, AI has no real reason to use it.

For example, a page that says:

“We provide professional air conditioning services in Melbourne. Our team is experienced, reliable, and affordable.”

That is not enough anymore.

Everyone says that.

AI does not need another generic service page.

But a page that says:

“We repair, service, and replace evaporative cooling, ducted heating, split systems, and gas heating units across Melbourne. Same-day service is available for urgent breakdowns. Our licensed technicians work with major brands including Braemar, Brivis, Bonaire, Fujitsu, Mitsubishi Electric, Daikin, Panasonic, and LG.”

That is much stronger.

It gives AI more context.

It gives users more confidence.

It gives Google more useful information.

It gives the business a better chance of being selected, summarised, and recommended.

SEO Is Becoming AI Visibility

This is the biggest mindset shift.

SEO used to mean:

How do we rank on Google?

Now it also means:

How do we get found, understood, and recommended by AI?

That includes:

• Google AI Overviews
• Google AI Mode
• ChatGPT
• Gemini
• Perplexity
• Copilot
• Voice assistants
• AI browsers
• On-device AI tools
• AI agents that research businesses for users

This is where the Chrome story becomes important.

If AI is being built into Chrome itself, then the browser becomes part of the discovery journey.

People may not just search websites anymore.

They may ask the browser:

“Find me the best heating and cooling company near me.”

“Summarise this business for me.”

“Compare these three plumbers.”

“Is this website trustworthy?”

“Does this company have good reviews?”

“Which electrician looks more reliable?”

“Can I book online?”

“Who offers same-day service?”

When that happens, your website needs to be machine-readable, trustworthy, structured, and useful.

Not just pretty.

What AI Will Look For On a Website

AI systems need clear information.

They do not care how much animation your homepage has.

They do not care that your hero section says “Your trusted local experts.”

They care about clear signals.

Things like:

• What services do you offer?
• Where do you offer them?
• Who do you help?
• What makes you different?
• Are you licensed?
• Are you insured?
• Do you have reviews?
• Do you show real contact details?
• Do you have pricing or quote information?
• Do you explain your process?
• Can users book or enquire easily?
• Is your content specific?
• Is your website technically clean?
• Is your business information consistent across the web?

This is especially important for local businesses.

Plumbers. Electricians. Heating and cooling companies. Builders. Lawyers. Dentists. Accountants. Real estate agencies. Medical clinics. Removalists. Cleaning companies. Auto repair businesses.

These businesses rely on search visibility.

But soon, they will also rely on AI visibility.

Local SEO Will Become Even More Important

AI does not remove local SEO.

It makes it more important.

When someone searches for a service near them, AI still needs local data.

It needs to know:

• Which businesses are nearby
• Which businesses are trusted
• Which businesses have strong reviews
• Which businesses clearly explain their services
• Which businesses have active websites
• Which businesses have consistent Google Business Profile information
• Which businesses are mentioned across directories and local sources
• Which businesses show real authority in their area

For example, if someone in Glen Waverley searches:

“Best split system installer near me”

AI will not just randomly pick a business.

It will look at signals.

Website content. Google Business Profile. Reviews. Service area pages. Local relevance. Business reputation. Content quality. Contact details. User experience.

This means local SEO cannot be lazy anymore.

Businesses need properly built service pages and suburb pages that actually help users.

Not duplicated pages where only the suburb name changes.

Generic Blog Content Will Not Be Enough

A lot of businesses still publish blogs like:

“5 Benefits of Air Conditioning”

“Why You Need a Plumber”

“Top Reasons to Hire an Electrician”

“What Is SEO?”

These topics are too generic.

AI can answer these without needing your website.

The better approach is to create specific, helpful, experience-based content.

For a heating and cooling company, better blog topics would be:

• How much does ducted heating repair cost in Melbourne?
• Should I repair or replace my evaporative cooler?
• Best heating options for older Melbourne homes
• Why does my gas ducted heating smell strange?
• How often should I service my split system in Victoria?
• Evaporative cooling vs split system cooling for Melbourne weather
• What causes carbon monoxide leaks in gas heaters?
• Is same-day heating repair available in winter?

For a plumber:

• What to do if your hot water system stops working overnight
• How much does blocked drain repair cost in Melbourne?
• Signs your stormwater drain is blocked
• When should you replace a hot water unit?
• Why your toilet keeps blocking
• Emergency plumber vs normal plumber: what is the difference?

For an SEO agency:

• How long does SEO take for a local business in Australia?
• Why your website ranks but does not generate leads
• SEO vs Google Ads for trades businesses
• What should be included in a monthly SEO report?
• Why suburb pages work when done properly
• How AI search is changing local SEO

This type of content is more useful.

It answers real questions.

It helps users make decisions.

And it gives AI systems better information to work with.

The New SEO Formula

The old SEO formula was:

Keywords + backlinks + content volume = rankings

The new SEO formula is becoming:

Helpful content + structured data + trust signals + brand authority + local relevance + technical quality = visibility across search and AI

This does not mean keywords are dead.

They are still important.

But keywords alone are not enough.

Your website needs to be built like a proper knowledge source.

A page should not just say what service you provide.

It should answer the questions a customer would ask before choosing you.

For example:

• What is the service?
• Who needs it?
• When should someone book it?
• How much does it usually cost?
• What affects the price?
• What suburbs do you service?
• What brands do you work with?
• What is your process?
• Why should someone trust you?
• What happens after someone submits an enquiry?

This is the content that performs better in an AI-driven search world.

Website Structure Matters More Than Ever

AI needs structure.

If your website is messy, vague, slow, or hard to crawl, you are making it harder for search engines and AI tools to understand your business.

Important things to fix:

• Clear heading structure
• Proper page titles and meta descriptions
• Schema markup
• FAQ sections
• Internal linking
• Fast page speed
• Mobile-friendly design
• Clear calls to action
• Clean navigation
• Service pages for each major offer
• Location pages where relevant
• Optimised Google Business Profile
• Consistent contact information
• Real reviews and testimonials
• Updated privacy policy
• Proper cookie consent if tracking users

These things may sound basic.

But many businesses still get them wrong.

And when AI starts summarising and comparing businesses, the websites with clean, clear, helpful information will have an advantage.

Privacy and Trust Will Become Part of the SEO Conversation

The Chrome article is mainly about consent and transparency.

That connects to SEO in a different way.

Users are becoming more aware of how their data is collected, tracked, analysed, and used by AI systems.

A business website that uses tracking pixels, cookies, analytics tools, AI chatbots, contact forms, booking systems, and email marketing platforms needs to be more transparent.

This does not directly mean privacy policies will become a ranking factor tomorrow.

But trust affects conversions.

And conversions are the real goal of SEO.

There is no point ranking if users do not trust the business.

Every business website should now review:

• Privacy policy
• Cookie banner
• Contact form consent
• CRM integrations
• Call tracking disclosure
• AI chatbot disclosure
• Email marketing opt-in process
• Third-party scripts
• Tracking pixels
• Data storage practices

This is especially important for industries like legal, healthcare, finance, real estate, and professional services.

If AI tools start reviewing websites on behalf of users, they may also become better at spotting trust issues.

A website with no privacy policy, no clear business details, poor reviews, broken pages, vague service information, and aggressive tracking will not look good.

Not to humans.

Not to AI.

The Browser Is Becoming a Gatekeeper

This is the biggest point.

Chrome is not just a browser anymore.

Google Search is not just a search engine anymore.

AI is becoming the layer between users and websites.

That means businesses need to think beyond traditional SEO.

In the future, your website may be judged before the user even lands on it.

AI may summarise your business. AI may compare you with competitors. AI may decide whether your content is useful. AI may decide whether your business looks trustworthy. AI may recommend you or ignore you.

This is why “just having a website” is not enough anymore.

Your website needs to act like a strong digital asset.

It needs to explain. It needs to prove. It needs to answer. It needs to build trust. It needs to convert. It needs to be readable by both humans and machines.

What Businesses Should Do Now

The businesses that prepare early will have an advantage.

Here is what I would focus on.

1. Update Your Core Service Pages

Every main service should have its own strong page.

Do not put everything on one generic services page.

If you are a heating and cooling company, you may need pages for:

• Ducted heating repairs
• Ducted heating servicing
• Gas heater repairs
• Evaporative cooling repairs
• Evaporative cooling servicing
• Split system installation
• Split system repairs
• Carbon monoxide testing
• Emergency heating and cooling repairs

Each page should clearly explain the service, who it is for, common problems, service areas, brands, process, FAQs, and call to action.

2. Build Real FAQ Content

FAQs are powerful because they match how people search.

They also match how people ask AI tools questions.

Good FAQs should answer real customer questions.

For example:

• How much does this service cost?
• How quickly can you attend?
• Do you service my suburb?
• Do you work with my brand?
• Is repair or replacement better?
• Do you offer emergency service?
• Are your technicians licensed?
• What happens after I submit an enquiry?

3. Add Schema Markup

Schema helps search engines understand your website.

Local businesses should consider:

• LocalBusiness schema
• Service schema
• FAQ schema
• Review schema where appropriate
• Breadcrumb schema
• Organisation schema

This does not magically guarantee rankings, but it improves clarity.

And clarity matters more in AI search.

4. Strengthen Google Business Profile

For local businesses, Google Business Profile is still one of the strongest visibility assets.

Businesses should keep it updated with:

• Correct services
• Correct categories
• Service areas
• Opening hours
• Photos
• Posts
• Reviews
• Review replies
• Booking links
• Website link
• Phone number

AI search will likely continue pulling local trust signals from Google’s own ecosystem.

So ignoring your Google Business Profile is a big mistake.

5. Create Better Blog Content

Stop publishing blogs just to tick a monthly SEO box.

Every blog should answer a real question or solve a real problem.

Better content will usually include:

• Specific examples
• Local context
• Pricing guidance
• Common mistakes
• Comparison tables
• Step-by-step explanations
• Expert advice
• FAQs
• Internal links to relevant services

The goal is not just to publish.

The goal is to become the best answer.

6. Improve Website Conversion

AI visibility is useful, but conversion still matters.

Once someone lands on your website, they should know exactly what to do.

Important conversion elements:

• Clear phone number
• Sticky call button on mobile
• Online booking or enquiry form
• Trust badges
• Reviews
• Before and after images
• Service area clarity
• Strong calls to action
• Fast loading pages
• Simple navigation

Many businesses do SEO to get more traffic, but their website leaks leads.

That problem will become worse if traffic becomes harder to win.

Every visitor matters more.

7. Be Transparent With Privacy and Tracking

This is where the Chrome article connects strongly.

As AI and tracking become more common, transparency matters.

Businesses should clearly explain:

• What data they collect
• Why they collect it
• What tools they use
• How contact form data is handled
• Whether AI chat tools are used
• Whether call tracking is used
• Whether third-party marketing tools are used

This is not just legal housekeeping.

It is part of brand trust.

What This Means for SEO Agencies

SEO agencies also need to adapt.

The old monthly deliverables model is not enough.

Clients do not just need “4 blogs and backlinks.”

They need:

• AI search readiness
• Website structure improvement
• Content strategy
• Conversion optimisation
• Local SEO
• Technical SEO
• Privacy and tracking review
• Schema implementation
• Google Business Profile optimisation
• Reputation strategy
• Better reporting around visibility, leads, and conversions

Agencies that keep selling old-school SEO without adapting to AI search will struggle.

Agencies that understand AI visibility, content quality, automation, and conversion strategy will win.

This is also a massive opportunity.

Because most businesses are not ready.

Most websites are still built for the old version of Google.

But users are moving into a new search experience.

What This Means for Business Owners

If you run a business, this is the simple version:

Your website needs to be clearer, more useful, and more trustworthy than before.

AI will reward businesses that explain themselves properly.

It will ignore businesses that hide behind generic marketing language.

You need to make sure your website clearly answers:

• What do you do?
• Who do you help?
• Where do you operate?
• Why should people trust you?
• What proof do you have?
• How can someone contact you?
• What happens next?
• What makes you different?
• What common questions do customers ask before buying?

If your website does not answer these clearly, you are making it harder for both people and AI to choose you.

The Big Picture

The Chrome 4GB AI model story is not only about a file sitting on someone’s computer.

It is about control.

Who controls the browsing experience? Who controls the search journey? Who decides what information users see first? Who decides whether websites get traffic or just become sources for AI answers? Who decides what gets installed, summarised, recommended, or ignored?

Google is clearly moving deeper into AI-powered browsing and search.

That means SEO is entering a new phase.

Not dead.

Not irrelevant.

But different.

The businesses that win will be the ones that understand this early.

They will not just chase rankings.

They will build websites that AI can understand, users can trust, and search engines can confidently recommend.

Final Thought

SEO is no longer just about being on Google.

It is about being part of the answer.

And in an AI-driven search world, the best answer will usually come from businesses that have done the hard work:

Clear content. Strong structure. Real trust signals. Local relevance. Technical quality. Helpful pages. Transparent data practices. Better user experience.

The Chrome AI model story shows where things are heading.

AI is not sitting outside the browser anymore.

It is moving into the browser, into search, into websites, into customer journeys, and into buying decisions.

So the question for businesses is not:

“Should we care about AI search?”

The better question is:

“When AI becomes the first layer between your customer and your website, will your business be ready?”

Right now, more than $2 trillion is being poured into AI.

This includes:

• Chips
• Data centres
• Cloud infrastructure
• Power plants
• AI start-ups
• Enterprise tools
• Corporate AI experiments

Everyone is racing to build the future. But there is one uncomfortable question nobody wants to ask too loudly:

What if the AI bubble bursts?

Not because AI is useless. But because the amount of money going into AI right now may be far ahead of the actual revenue, adoption, and return on investment it is producing. And if that gap gets exposed, the impact will not stay inside Silicon Valley. It could affect the stock market, retirement accounts, pension funds, energy prices, jobs, small businesses, and the broader economy.

The Scary Part Is Most People Are Already Exposed

This is not only a problem for venture capitalists and big tech.

A lot of people think the AI bubble is only a problem for venture capitalists, tech founders, and big companies.

But that is not really true.

If you have money in an index fund, super fund, pension fund, or retirement account, there is a good chance you are indirectly exposed to AI-heavy companies.

Today, investing in broad market indexes often means betting heavily on a small group of mega-cap technology companies.

Companies like Nvidia, Microsoft, Alphabet, Amazon, Meta and others are not just tech companies anymore.

They are now major AI infrastructure bets.

So when people put money into the market, a portion of that money is flowing into the AI boom whether they realise it or not.

That is why this matters.

If AI stocks correct sharply, the impact does not only hit billionaire investors.

It hits normal workers, retirees, teachers, nurses, public employees, small investors, and families who thought they were just investing safely for the long term.

The AI Economy Is Being Held Up By Massive Circular Spending

One of the biggest concerns is something called circular funding.

This is when major AI companies, cloud companies, chip companies, and infrastructure providers invest in each other, buy from each other, and create revenue loops that make the market look stronger than it may actually be.

For example,

• A chip company invests in an AI company.
• The AI company then uses that money to buy chips from the same chip company.
• On paper, both sides look stronger.
• The chip company shows revenue growth.
• The AI company looks well-funded.
• Investors get excited.
• Valuations go up.

But the question is:

How much of this demand is coming from real customers paying for real value?

That is the key issue.

Because if the money is mostly moving around between the same group of companies, then the market may be confusing circular spending for genuine economic demand.

That can work for a while.

But it becomes dangerous when everyone eventually needs real cash, real revenue, and real profit.

The Cost Of Running AI Is Still Extremely High

AI companies are growing fast, but many of them are also burning huge amounts of money.

Training and running large AI models requires massive computing power.

That means expensive chips, cloud infrastructure, electricity, cooling, data centres, and engineering teams.

So even when AI companies generate large revenue numbers, the cost of delivering the product can still be enormous.

This creates a strange problem:

The more people use the product, the more expensive it becomes to run.

That is very different from traditional software, where once the product is built, adding more users can become highly profitable.

With AI, every prompt, image, video, workflow, or agent task can create real compute costs.

So the business model depends on future improvements.

AI companies are betting that compute will become cheaper, models will become more efficient, and revenue will eventually grow faster than costs.

That may happen.

But it has not been fully proven yet.

The ROI Problem Is The Biggest Red Flag

The biggest question for businesses is simple.

Is AI actually producing measurable return on investment?

Many companies are testing AI.

Many teams are using ChatGPT, Copilot, Claude, Gemini, automation tools, AI agents, and AI content platforms. But using AI is not the same as generating profit from AI. A lot of businesses are still using AI in a very manual way.

• They ask AI to create something.
• Then a human checks it.
• Then a human edits it.
• Then a human implements it.

So yes, AI may speed up parts of the work.

But in many cases, the human is still doing the final work.

That means the productivity gain is real, but not always transformational.

And that is where the problem starts.

If companies spend billions on AI but cannot prove measurable returns, investors will eventually ask:

Where is the profit?

When that question becomes louder, valuations can change very quickly.

If The Bubble Bursts, The Stock Market Gets Hit First

The first impact would likely be felt in the stock market.

AI-related companies have contributed heavily to recent market growth.

If investor confidence shifts, AI stocks could correct sharply.

That could drag down broader indexes because so much market performance is concentrated in a small number of technology companies.

This could affect:

• Stock portfolios
• Retirement accounts
• Super funds
• Pension funds
• ETFs
• Tech-heavy index funds
• Start-up valuations
• Investor confidence

The damage could be psychological as much as financial.

Once investors stop believing that AI will deliver huge profits quickly, the market will reprice the entire sector. And when the market reprices hype, it usually does not do it gently.

The Second Risk Is Debt

The AI buildout may be different because it is not just funded by equity.

The dot-com crash was mostly an equity story. A lot of internet companies were overvalued, many collapsed, and investors lost money.

But the AI buildout may be different because it is not just funded by equity.

It also involves debt;

• Data centres are expensive.
• Power infrastructure is expensive.
• Chip purchases are expensive.
• Cloud commitments are expensive.

Corporate bonds, private credit, joint ventures, and off-balance-sheet financing are increasingly being used to fund the AI race.

That makes the risk more serious.

If the bubble bursts and the debt behind the infrastructure becomes difficult to service, the problem could move beyond tech stocks.

It could affect lenders, banks, private credit markets, energy contracts, commercial property, construction, and infrastructure companies.

That is when the situation starts to look less like the dot-com crash and more like a wider financial stress event.

Jobs Will Be Caught In The Middle

The strange thing about AI is that both success and failure can hurt jobs.

If AI works extremely well, companies may use it to reduce headcount.

If AI fails to deliver returns, companies may cut jobs to protect margins.

That means workers can get squeezed from both sides.

In a full AI bubble burst scenario, the job losses would likely start in tech and then spread into related sectors.

The most exposed areas could include:

• AI start-ups
• SaaS companies
• Cloud consulting
• Data centre construction
• Chip supply chains
• Recruitment
• Customer support roles
• Admin-heavy roles
• Junior white-collar roles
• Software teams with duplicated functions

But the impact would not be equal.

Some jobs would disappear.

Some would be reshaped.

Some would become AI-assisted.

And some new roles would appear.

The real risk is not that every job disappears overnight.

The real risk is a messy transition where companies cut faster than workers can adapt.

Data Centres Could Become The Physical Symbol Of The Bubble

The AI boom is not just software.

It is physical.

Data centres need land, power, water, cooling, chips, fibre connections, and long-term energy agreements.

Big tech companies are making enormous commitments to power AI infrastructure.

• Some are signing energy deals.
• Some are supporting the reopening or expansion of power plants.
• Some are building massive campuses that require huge amounts of electricity.

This creates a public risk.

If the AI revenue does not justify the infrastructure, the contracts and costs do not simply disappear.

Someone still has to pay.

That could mean higher electricity costs, government incentives, tax concessions, local infrastructure pressure, and long-term costs passed on to households and businesses.

In other words, the private upside goes to tech companies.

But if things go wrong, some of the cost may be socialised.

That is the part people should pay attention to.

This Is Different From The Dot-Com Crash

A lot of people compare AI to the dot-com bubble.

That comparison is useful, but not perfect.

When the dot-com bubble burst, it left behind valuable infrastructure.

Fibre optic cables were laid during the boom.

After the crash, that infrastructure became cheaper and helped support the next generation of internet companies.

Google, YouTube, Netflix, cloud platforms, and many other digital businesses benefited from the infrastructure built during the hype cycle.

AI infrastructure may not work the same way.

The biggest issue is chips.

AI chips can lose economic value quickly.

They are expensive, they age, and new generations of hardware can make older chips less competitive.

A data centre full of outdated GPUs is not the same as fibre optic cable that can keep producing value for decades.

So if the AI buildout overshoots demand, some of the infrastructure may not become a long-term bargain.

It may become expensive scrap.

How The Bubble Could Deflate

The AI bubble probably would not burst in one dramatic day.

It would likely deflate over months or years.

The likely sequence could look like this:

• First, investors start questioning revenue quality.
• Then start-ups struggle to raise funding.
• Then companies cut AI pilots that do not show ROI.
• Then weaker AI tools get commoditised by bigger model providers.
• Then funding dries up.
• Then valuations fall.
• Then layoffs increase.
• Then data centre projects are paused or cancelled.
• Then the strongest companies buy the useful assets from the weaker ones.

This is how hype cycles usually unwind.

Slowly at first.

Then suddenly.

The Big Winners May Become Even Bigger

If the AI bubble bursts, AI probably does not disappear.

Instead, it may consolidate.

• Smaller AI start-ups may fail or get acquired.
• AI wrapper tools may disappear.
• Weak products with no moat may be crushed.
• The biggest companies with cash, cloud infrastructure, distribution, and existing customers may survive.

That means Microsoft, Google, Amazon, Meta, Nvidia, and a few others could end up controlling even more of the AI economy.

So the crash may not end AI.

It may simply reduce the number of players.

The AI revolution could continue, but in a slower, more practical, and more concentrated way.

The Hype Dies, But The Useful Technology Stays

This is the part I think people need to understand.

AI is not fake. But some AI valuations probably are.

AI will continue to be useful in many areas:

• Customer support
• Sales follow-up
• Workflow automation
• Coding assistance
• Research
• Admin tasks
• Document summarisation
• Internal knowledge search
• Marketing support
• Data analysis
• Business process automation

The real value will come from boring, practical use cases.

The companies that win will be the ones that use AI to reduce cost, improve speed, increase revenue, and make teams more productive.

That is where AI becomes real.

What Businesses Should Do Now

For business owners, the lesson is not to ignore AI.

The lesson is to avoid blindly chasing it. Use AI where it creates measurable value.

Ask simple questions:

• Does this reduce manual work?
• Does this save time?
• Does this improve customer experience?
• Does this help us respond to leads faster?
• Does this reduce admin cost?
• Does this increase conversion?
• Does this improve delivery?
• Does this give the team more leverage?

If the answer is yes, AI is worth exploring.

If the answer is no, it is probably just hype.

Businesses do not need AI theatre. They need AI that actually improves operations.

My Take

I do not think AI is going away.

But I do think the market may eventually separate real AI value from AI hype.

• A lot of weak companies will disappear.
• A lot of overvalued start-ups will struggle.
• A lot of AI tools will become features inside bigger platforms.
• A lot of investors will realise that not every AI company deserves a billion-dollar valuation.

But the useful parts of AI will survive.

Just like the internet survived the dot-com crash.

The bubble may burst.

The technology may stay.

And the people who benefit most will be the ones who understand the difference between hype and utility.

Because in the end, technology does not win because it is exciting.

It wins because it becomes useful.

If you’ve been using AI to code, you’ve probably run into this loop:

• You give it a prompt → it writes something → something breaks → you fix it → prompt again

Repeat that a few times and eventually you get something working.

It’s fine for small things. But once the project gets even slightly complex, it starts to fall apart.

You end up spending more time fixing, re-prompting, wasting tokens and double-checking than actually building.

I tested this recently: https://github.com/obra/superpowers

And it changed how I think about using AI for development.

The issue isn’t the AI

Most people blame the model.

• “It’s not smart enough”
• “It missed this logic”
• “It didn’t follow instructions”

But after using this, it’s pretty clear that’s not the real problem.

The issue is how we’re using it.

We’re asking AI to jump straight into code without doing any of the thinking that normally happens before development.

• No clear problem definition
• No structured breakdown
• No planning

Just “build this” and hope it gets it right.

That’s why things feel inconsistent.

What this does differently

Instead of jumping straight into execution, this framework forces a proper flow.

Before any code is written, it goes through:

• Understanding the problem
• Designing the solution
• Breaking it into steps
• Then executing

It sounds basic, but most people skip this when using AI.

And that’s exactly why they get messy results.

Once you introduce structure, the output becomes a lot more predictable.

What I actually noticed

The biggest difference wasn’t that the AI became “smarter”

It’s that it became more controlled.

It started behaving less like a tool you constantly correct And more like something you guide

It also enforces a few things that people usually ignore when working with AI:

• Writing tests before jumping into implementation
• Debugging step by step instead of randomly patching issues
• Reviewing outputs instead of assuming they’re correct
• Working in smaller chunks instead of trying to solve everything in one go

These are all things experienced developers already do.

But when using AI, most people skip them.

Why this matters (especially if you’re learning with AI)

If you’re using AI to learn coding, it’s very easy to fall into the habit of just prompting your way through problems.

And it feels productive at first.

But over time, it actually slows you down.

Because you’re not building a way of thinking You’re just reacting to outputs

What this kind of approach does is force you to think properly before execution.

You start asking better questions:

• Did I define the problem clearly?
• Did I break this down properly?
• Am I guiding the AI step by step?

That’s where real improvement happens.

The bigger shift

What I’m starting to see is a shift in how people build with AI.

It’s moving from:

• Prompt → Code

To something more structured:

• Process → System → Outcome

And this matters more than people think.

Because the ones getting real results with AI aren’t the ones writing clever prompts.

They’re the ones building repeatable systems around it.

Final thought

AI isn’t here to replace developers.

But it is changing how development works.

If anything, it’s making structured thinking more important.

The better your process, the better your results.

If you’re building with AI, this is definitely worth exploring: https://github.com/obra/superpowers

You’re running Claude in one terminal. Codex in another. Gemini in a third.

You read the output from one.

• Copy it.
• Switch tabs.
• Paste it into the next.
• Explain the context again.
• Wait.
• Then switch back.

If you’re using multiple AI tools right now, you’ve probably realised something strange.

You are basically acting as the middleware between them.

Developers have spent years removing manual integration work but somehow with AI tools, we’re doing it again. Manually passing information between agents that don’t know the others exist. And while you’re prompting one AI, the others are just sitting there idle.

You might have three AI assistants open, but you’re still only able to use one at a time.

That’s the problem Colana is trying to solve.

One Command Changes the Workflow

npm install -g colana

No setup.

No config.

No account creation.

You install it and it opens a local control centre for your AI agents. Instead of jumping between terminals, everything runs in one place.

What Colana Actually Does

Colana lets you run multiple AI coding agents side by side in real terminal sessions.

• Not wrappers.
• Not API layers.
• Actual interactive terminals.

You can run tools like Claude, Codex, Gemini, Copilot CLI, and Aider at the same time.

Each agent has its own terminal pane, so you can see what they’re doing and interact with them directly.

There’s also shared project context, so agents don’t need you constantly copying and pasting information between them.

Colana also isolates git changes automatically when multiple agents are working on the same repository, which prevents file conflicts.

And every session is saved, searchable, and resumable later.

What This Looks Like in Practice

Here are a few simple ways people are using it already.

Running agents in parallel

• Claude writes a feature.
• Gemini reviews it.

Both are running at the same time on the same screen.

You’re not waiting for one to finish before starting the next.

Comparing AI models

Give two agents the same task and run them side by side.

You can instantly see which model gives better results for your workflow instead of guessing.

Assigning work to multiple agents

Instead of chatting with one AI at a time, you can assign tasks to several.

• Give Codex one job.
• Switch to Claude and give it another.

Both keep working while you move between them.

It feels less like using a single AI assistant and more like managing a small AI team.

Why This Matters

The big shift happening in AI right now isn’t just better models. It’s how workflows are changing. We’re moving from:

one AI assistant

to

multiple specialised AI agents working together.

When that happens, the real challenge becomes coordination. Tools like Colana are starting to solve that layer.

Early traction

Colana already hit 12,000 installs in its first five days.

No launch campaign.

No ads.

Just developers sharing it with each other.

Pricing

The tool is free forever for two concurrent agents.

All core features included.

• No trial.
• No credit card.

Free during beta. Full access. Everything unlocked.

No account, no email, no phone number, and no personal data collected. Just install and code.

Try it

npm install -g colana

You might quickly realise how much time you’ve been spending alt-tabbing between AI tools.

Explore Colana

🌐 https://colana.ai 📦 https://www.npmjs.com/package/colana

Over the past year, almost every AI product I built relied on OpenAI.

It made sense. Best models. Simple API. Fast to ship.

But recently, I’ve started moving parts of my stack towards open-source AI infrastructure.

Not because OpenAI isn’t powerful.

But because depending entirely on one closed AI provider is a long-term business risk.

If you’re building SaaS, AI agents, automation tools, or chatbots, this matters.

Here’s why.

1. Vendor Lock-In Is a Real Risk in AI

If your entire app depends on a closed AI API like OpenAI:

• They can increase pricing anytime
• Change rate limits
• Modify model behaviour
• Deprecate or discontinue models

You don’t control your own foundation.

Your product roadmap becomes dependent on their roadmap.

With open-source AI models, you can:

• Switch providers instantly
• Self-host models if needed
• Avoid platform dependency
• Build multi-provider redundancy

This gives you long-term independence.

For serious founders, that flexibility is critical.

2. Open-Source AI APIs Reduce Costs at Scale

When you’re small, API costs don’t hurt much.

When you scale, they become one of your biggest operational expenses.

Closed AI APIs include:

• Model training costs
• Infrastructure overhead
• Profit margins
• Investor pressure

Open-source AI APIs and inference providers are significantly cheaper.

If you’re building:

• SaaS platforms
• AI agents
• Customer support automation
• AI-powered workflows
• Chatbots

Cost efficiency directly impacts your margins.

At scale, this becomes the difference between profitable and unprofitable AI products.

3. Better Data Privacy and Infrastructure Control

With closed AI APIs:

• Your data goes to external servers
• Logging depends on provider settings
• You don’t control infrastructure

For enterprise-grade applications, this is a major consideration.

Open-source AI infrastructure allows:

• Full self-hosting
• Private deployment
• Complete data ownership
• Custom optimization

If you’re targeting enterprise clients, this control becomes a competitive advantage.

4. Open-Source AI Models Are Now Extremely Competitive

Two years ago, open-source AI models were far behind.

Today, that’s no longer true.

Modern models like:

• DeepSeek
• Llama
• Mixtral
• Qwen

Are highly capable for most real-world SaaS use cases.

Including:

• Customer support
• AI agents
• Internal tools
• Summarization
• Workflow automation

For many applications, you don’t need frontier-level closed models.

Open-source AI is already more than sufficient.

Where OpenAI Is Weak Right Now

This isn’t criticism. It’s business reality.

OpenAI helped accelerate the entire AI ecosystem.

But there are risks founders should understand.

1. Financial Sustainability Uncertainty

Running frontier AI models costs billions.

Which means:

• Pricing structures will evolve
• Limits will change
• Monetization strategies will shift

If your entire business depends on one provider’s economics, that’s strategic risk.

2. Limited Infrastructure Control

With closed AI APIs, you cannot:

• Modify core model behaviour
• Run models locally
• Fully optimize inference performance
• Control long-term infrastructure decisions

You are dependent on their system.

3. Strategic Platform Risk

They can:

• Change pricing
• Adjust rate limits
• Prioritize enterprise customers
• Shift product direction

And you have to adapt.

That’s not ideal if your entire SaaS relies on them.

4. Reliability and Business Continuity Risk

If your AI provider goes down, your product goes down.

Their outage becomes your outage.

Open-source AI infrastructure allows multi-provider redundancy and optional self-hosting.

That reduces operational risk significantly.

The Smart Approach: Hybrid AI Infrastructure

The goal isn’t to avoid OpenAI completely.

The goal is to avoid total dependency.

A strong AI architecture in 2026 looks like this:

• Primary layer → Open-source models via OpenRouter or Together AI
• Primary models → DeepSeek or Mixtral
• Fallback → OpenAI
• Optional → Self-host critical workloads

This gives you:

• Redundancy
• Cost control
• Flexibility
• Scalability
• Long-term protection

Closed AI APIs like OpenAI are powerful.

But building your entire product on one provider creates unnecessary risk.

Open-source AI infrastructure gives you:

• Lower operational costs
• Infrastructure independence
• Greater control
• Enterprise-grade flexibility
• Long-term scalability

The smartest founders today aren’t anti-OpenAI.

They’re just not dependent on it.

If you’re building an AI product and want to design a scalable, cost-efficient AI stack, happy to chat.

In late January 2026, an open-source project called OpenClaw exploded across GitHub, developer forums, and social media. What started as a small Node.js experiment quickly crossed 100,000 GitHub stars and attracted millions of curious engineers, founders, and tinkerers.

The reason was simple: OpenClaw promised something most AI tools still do not deliver. It does not just answer questions. It takes action.

Built by Peter Steinberger, OpenClaw was designed as a personal AI agent that lives alongside your digital life. Instead of chatting inside a browser tab, it connects large language models like Claude and GPT directly to your messaging apps, files, calendar, and operating system. You can message it to clean your inbox, reschedule meetings, run scripts, or manage tasks and it actually does the work.

This article breaks down how OpenClaw works, why it captured global attention, what makes it different from traditional AI tools, and the very real security concerns that come with autonomous AI agents.

From Clawdbot to OpenClaw: A Fast and Messy Evolution

OpenClaw did not start with its current name. The project was originally released as Clawdbot, before briefly becoming Moltbot, and finally settling on OpenClaw within a matter of days.

The rapid rebranding was triggered by trademark concerns and growing visibility. Each name change expanded the audience and amplified scrutiny. As excitement grew, so did unintended consequences. Security teams began noticing employees installing and using OpenClaw variants without approval, turning it into a textbook example of shadow IT.

The rebrands also created confusion that scammers were quick to exploit. Look-alike repositories, fake installers, and malicious clones appeared almost immediately. The lobster mascot and playful branding helped the project spread, but the speed of adoption outpaced many users’ understanding of the risks.

How OpenClaw Actually Works

At its core, OpenClaw is not an AI model. It is an orchestration layer that connects existing language models to real-world tools.

The system is built around four core components:

• Gateway A background service that stays online and listens for messages. It connects OpenClaw to messaging platforms like WhatsApp, Telegram, Discord, Slack, and others.
• Agent The reasoning layer powered by a large language model. This component interprets your request, plans the steps, and decides which tools to use.
• Skills Modular capabilities that give the agent access to browsers, files, shell commands, calendars, APIs, and third-party services. Skills are what allow OpenClaw to move from conversation to execution.
• Memory Persistent storage that saves context across sessions. OpenClaw remembers preferences, ongoing projects, and past decisions so interactions improve over time.

Because the Gateway runs continuously, OpenClaw behaves more like a service than a traditional app. It keeps working even when your laptop is closed. This is why many users deploy it on a server or VPS rather than treating it like a normal desktop tool.

What Makes OpenClaw Different

OpenClaw stands out because it combines several capabilities rarely found in one system.

• Chat-first experience You interact with OpenClaw through everyday messaging apps. There is no special interface to learn. It feels like messaging a coworker rather than using software.
• Deep system access OpenClaw can open browsers, click through websites, read and write files, run shell commands, schedule jobs, and interact with connected devices. This level of access is what allows it to complete tasks end-to-end.
• Persistent memory Unlike most AI tools that forget everything after a session, OpenClaw remembers. This turns it from a chatbot into a long-term assistant that adapts to how you work.
• Extensible skills ecosystem The community is rapidly building skills that expand what OpenClaw can do. In some cases, the agent can even create and install new skills on its own.
• Proactive behaviour OpenClaw does not just wait for commands. It can run recurring tasks, monitor conditions, and surface suggestions when something needs attention.
• Model-agnostic architecture Although many users prefer Claude for reasoning, OpenClaw works with multiple AI providers and even local models. Switching models does not require rebuilding the system.
• Self-hosted and privacy-friendly Because it runs on your machine or server, your data stays under your control. This is a major reason privacy-conscious developers are drawn to it.

Setup and Deployment

Despite viral videos showing single-line installs, OpenClaw is not a plug-and-play consumer app. Setup is straightforward but requires care.

The typical process involves installing the core system, running an onboarding wizard, configuring AI provider keys, connecting messaging platforms, and testing commands. Advanced users then customize memory behaviour, install skills, enable sandboxing, or deploy OpenClaw on a cloud server for always-on access.

One important misconception is hardware requirements. OpenClaw does not need expensive machines. Many users successfully run it on low-cost VPS instances with modest CPU and memory. The heavy computation happens on the AI provider’s side. Locally, OpenClaw is coordinating actions rather than crunching data.

The creator has repeatedly advised users not to buy dedicated hardware just for OpenClaw. Cheap servers or existing machines are more than sufficient.

Real-World Use Cases

The excitement around OpenClaw comes from how quickly it proves useful in everyday scenarios.

• Productivity automation Users rely on it to manage email, draft replies, schedule meetings, summarize inboxes, and send daily briefings.
• Developer assistance OpenClaw can run tests, debug issues, monitor repositories, deploy updates, and even open pull requests. For engineers, it feels like a junior teammate that never sleeps.
• Personal operations From booking travel to tracking expenses and generating reports, OpenClaw handles tasks that usually require switching between multiple apps.
• Smart-home control When connected to home automation systems, it can manage lighting, temperature, and devices through natural language commands.
• Creative workflows Some users use OpenClaw to edit images, generate content, and build dashboards by describing what they want instead of manually assembling tools.

This breadth of use is why many describe OpenClaw as a glimpse of the future rather than just another AI tool.

The Security Reality No One Should Ignore

OpenClaw’s power is also its biggest risk.

Because it has persistent memory and deep system access, mistakes carry real consequences.

• Plain-text memory storage Configuration files and long-term memory are stored in readable formats. If the host machine is compromised, attackers can access sensitive context, credentials, and logs instantly.
• High privilege levels OpenClaw often runs with permissions comparable to an administrator. A misunderstood instruction or malicious prompt can trigger destructive actions.
• Exposed dashboards Researchers have found many OpenClaw instances accessible on the public internet due to misconfiguration. These exposures revealed chat histories, API keys, and remote command access.
• Prompt injection attacks If OpenClaw reads untrusted content such as emails, documents, or web pages, attackers can embed instructions that manipulate the agent into leaking data or executing commands.
• Malicious skills Third-party skills can introduce hidden behaviour. Some skills have been shown to silently exfiltrate data or bypass safeguards.
• Enterprise blind spots When employees run autonomous agents on personal machines or VPSs, traditional security tools often cannot see or control what the agent is doing.

These are not hypothetical risks. They are already happening.

How to Use OpenClaw Safely

OpenClaw can be explored responsibly if treated with the seriousness it deserves.

Run it in isolation on a dedicated server or virtual machine Avoid exposing dashboards directly to the public internet Use sandboxing and restrict write access to sensitive directories Install skills only from trusted sources and review code Separate the agent’s identity from your personal accounts Rotate credentials regularly and monitor logs Stay up to date with security improvements and patches

The project is actively improving its security posture, but the responsibility still sits with the user.

Final Thoughts

OpenClaw proves that agentic AI is no longer theoretical. It shows how language models, tools, and memory can combine into assistants that genuinely act on our behalf.

At the same time, it exposes how unprepared most individuals and organisations are for autonomous software with real authority. The future OpenClaw points to is powerful, productive, and deeply disruptive. It is also risky if adopted without discipline.

For builders and early adopters, OpenClaw is an exciting laboratory. For businesses, it is a warning sign. The next generation of AI will not live behind dashboards and approval flows. It will live in chats, servers, and background processes.

The question is no longer whether AI agents will arrive. It is whether we will be ready when they do.