Shoeb Ali | NEXORA.Crypto

Last week: the Bitcoin arms race between nations. This week: what happens when the financial weapons those nations have relied on for decades start to crack.

The global financial system was built on one assumption: centralised infrastructure can be controlled. Cryptocurrency breaks that at a fundamental level. In 2025, sanctioned economies are using decentralised networks to maintain trade that traditional tools were designed to sever. Nation-states are racing to deploy their own digital currencies in response. And a multi-decade de-dollarization trend is accelerating, with crypto at the centre. For developers, this is not background noise. The code you write right now operates inside a geopolitical context that didn’t exist five years ago.

Picture the moment a finance minister realises the sanctions meant to cripple their economy, the ones that worked reliably for decades, aren’t working the way they used to.

The playbook relied on one thing: the global financial system being centralised. Banks, clearinghouses, payment processors. All identifiable choke points, switchable off with a call to the right regulator.

Now that same sanctioned government has been quietly routing significant trade through decentralized blockchain networks for two years. Not through banks. Through networks no single authority controls.

That’s not a hypothetical. That’s the world we’re already in.

What happens to global power when the primary tool of financial coercion stops working?

The answer reshapes everything, including the architecture of the code you’re building today.

(First time here? This stands alone. But last week’s piece on the Bitcoin arms race is worth 7 minutes.)

The Architecture Problem Nobody Wants to Admit

Cryptocurrency didn’t just create a new asset class. It created a financial infrastructure layer designed to resist centralised control. That’s not a bug. It’s the feature.

When sanctioned economies get cut off from conventional banking, crypto becomes survival infrastructure, not speculation. A single exchange operating in the sanctions zone of one affected economy processed an estimated $9 billion in transactions in 2025 alone. Blockchain analytics firms tracked over a billion dollars in state-linked theft across 2024–2025. Entire alternative payment ecosystems are being built to settle trade independent of dollar-dominated rails.

These aren’t workarounds. They’re a sign of how financial coercion has changed, or increasingly, stopped working the way it once did.

One distinction worth holding: crypto transactions aren’t anonymous, they’re pseudonymous. Every transaction is permanently recorded on a public ledger. The problem isn’t visibility; it’s that enforcement infrastructure is still racing to catch up with the actors exploiting the gaps.

This is Article 2 of a 6-part series. If the geopolitical angle is landing, hit Follow and Subscribe above. Next week: how Wall Street and DeFi are quietly merging in ways that will reshape how everyone in this space builds and invests.

The State’s Counter-Move: CBDCs

Governments don’t give up financial control without a fight. In 2025, the most significant counter-move isn’t just regulation. It’s competition.

Central Bank Digital Currencies (CBDCs) are the nation-state’s direct answer to decentralised crypto. As of mid-2025, more than 130 countries, representing over 98% of global GDP, are actively researching, piloting, or deploying them, per the Atlantic Council’s tracker.

Here’s the thing about CBDCs that most people miss: they’re not digital cash. They’re programmable money. Expiry dates that force spending. Spending restrictions. Automatic tax withholding baked into every transaction. Geographic limitations that make it unusable outside certain regions.

That creates a fork every developer building financial infrastructure will eventually have to navigate:

Decentralised crypto: permissionless, censorship-resistant, privacy-preserving by design. Hard to control. Hard to surveil.

CBDCs: efficient, interoperable, and fully traceable by design. Easy to control. Easy to surveil.

Neither is inherently evil. But they represent very different ideas about what money is for and who gets to see it. The infrastructure you choose to build for, or bridge between, reflects what you think money should do. That’s not a small choice.

What This Means for Developers

The geopolitical context shapes every technical decision you’ll make here. Here’s what that looks like in practice:

State-level adversaries are real, not theoretical. In early 2025, blockchain analysts attributed a $1.5 billion theft from a major centralised exchange to a state-sponsored operation. It was the single largest crypto theft in history at that point. Separately, a major exchange in a sanctions-affected region saw wallet balances collapse from approximately $1.8 billion to near $100 million in hours. For anyone building exchange infrastructure, custodial solutions, or wallet systems: build your threat models accordingly, today.

Compliance is architecture, not a checkbox. Sanctions compliance means understanding how your code interacts with international financial infrastructure. On-chain analytics are sophisticated enough now that “I didn’t know” is an increasingly thin defence.

Privacy is a policy fight you’re already inside. The tension between crypto’s privacy architecture and governments’ demand for financial transparency will shape regulation for a decade. Build with a clear understanding of both sides, not just the engineering.

CBDC interoperability is real demand, but surveillance is built in from day one. The opportunity is genuine. But this infrastructure won’t have surveillance added later as an afterthought. It will be designed in from the start. Know where you stand on that before you sign the contract.

The Road Ahead

CBDCs will spread. Regulatory frameworks will diverge sharply by jurisdiction, some built to enable, some to restrict, most to surveil. And the de-dollarization trend is generational. It will keep driving demand for assets like Bitcoin, which no country controls, no central bank can inflate, and no sanctions regime can freeze.

Crypto has moved past the “interesting technology” phase into something harder: infrastructure that governments and institutions are actively fighting to shape. The choices you make about privacy, compliance, and governance aren’t just engineering decisions. They’re positions in an ongoing argument about who controls money and who doesn’t.

It’s a remarkable moment to be building in. Stay awake to it.

Before you scroll: Where are you building right now: decentralised rails, CBDC infrastructure, or the bridge layer between them? What’s shaping that decision?

Key Takeaways

• Cryptocurrency has become a genuine geopolitical instrument. Decentralized infrastructure creates challenges that traditional financial coercion tools simply were not designed to handle
• CBDCs are the state’s counter-move: programmable, traceable digital money that preserves government control while competing with crypto on efficiency
• Bitcoin has a property no fiat currency can match: no country controls it, no central bank can print more of it, and no sanctions regime can freeze it. That makes it a serious option for reserve diversification
• Developers need to take geopolitical context seriously. State-level threat models, compliance complexity, and CBDC surveillance architecture are this cycle’s defining engineering challenges

#Cryptocurrency #Decentralized #Bitcoin

Shoeb Ali | NEXORA.Crypto | Newsletter Series 1 of 6

Nations are quietly stockpiling Bitcoin as a geopolitical weapon — and the window to understand why is closing faster than almost anyone realises

One major world power holds over 200,000 Bitcoin.

Its primary geopolitical rival holds nearly as many.

Nations under economic pressure have legalized crypto to maintain trade flows outside traditional banking systems.

And most people are still debating whether Bitcoin is “real money.”

The arms race is already underway. We just haven’t announced a winner yet — and the window for everyone else to respond is closing faster than almost anyone realises.

In 2025 alone, at least one major economy formalized its Strategic Bitcoin Reserve through executive order, over a dozen nations are actively exploring or implementing similar programs, and institutional concentration has reached levels that would have seemed impossible five years ago. This isn’t speculation — it’s a structural shift in how countries think about monetary sovereignty. For developers and crypto-curious readers alike, the infrastructure being built right now sits at the intersection of the most significant geopolitical transformation since the gold standard. Understanding this dynamic isn’t optional — it’s essential.

Picture a finance minister in a small nation watching the world’s largest economies trade Bitcoin like geopolitical poker chips. The treasury team runs the numbers: if just 1% of global foreign reserves shifted into Bitcoin, that’s roughly $80 billion in new demand for an asset that produces only about 450 new BTC per day. The math is simple, the signal is clear, and the window to act is narrowing. This isn’t hypothetical. This is happening right now, and the implications stretch far beyond price.

For the Crypto Enthusiast — The Validation You've Been Waiting For     

If you've been holding Bitcoin, you just got the most powerful signal in 
the asset's history — and most people still haven't processed it.       

Nation-states don't accumulate assets they plan to abandon. When a 
government formalizes a Bitcoin reserve, it signals something deeper 
than price confidence — it signals that Bitcoin has crossed the threshold 
from speculative instrument to strategic necessity. For anyone who 
understood this thesis years ago, the validation is real. But it also 
changes the game fundamentally: the asset you hold is now competing for 
supply with entities that have printing presses and no position limits. 
The scarcity argument you've always made just became a geopolitical 
argument backed by sovereign treasuries. That's a different conversation 
entirely.

How We Got Here: From Seized Assets to State Treasuries

Bitcoin’s journey from a cypherpunk experiment to a sovereign asset class reads like a geopolitical thriller. In 2013, a federal law enforcement agency auctioned 30,000 BTC seized from the Silk Road darknet marketplace — a novelty sale that attracted mostly individual bidders. Fast forward to March 2025, and a formal executive order established the first official national Strategic Bitcoin Reserve (SBR), built from approximately 200,000 BTC already held by the government from law enforcement seizures. The change in just over a decade is staggering.

The pioneer is not alone. According to Glassnode and Gemini’s 2025 Trends Report, 216 institutional entities now hold over 30% of all circulating Bitcoin. That concentration wasn’t built by retail investors in Discord servers.

Nation-states, corporate treasuries, and ETF custodians are now the dominant force in Bitcoin’s supply architecture. Nations across South Asia, Central America, Central Asia, and Western Europe have either accumulated Bitcoin or debated doing so publicly. Several advanced Scandinavian and alpine economies are actively studying integration into national reserves.

A second major world power maintains a characteristically opaque position — but the numbers speak clearly. Estimates suggest its state-associated wallets hold approximately 194,000 BTC, largely accumulated through a massive 2019 Ponzi scheme seizure. The official domestic stance remains anti-crypto, yet these holdings tell a different story — strategic preparation for a post-dollar financial world.

For the Crypto Enthusiast — The Supply Story Nobody Is Talking About

216 institutions now hold 30% of all Bitcoin. That supply isn't coming 
back to the market.

For anyone tracking on-chain fundamentals, this concentration matters far 
beyond price. Freely circulating Bitcoin — the coins actually available on 
exchanges — is structurally smaller than the headline numbers suggest. 
Sovereign and institutional holders treat their BTC as a long-term reserve 
asset, not a trading position. That means fewer coins on order books, 
tighter liquidity, and a market where each new wave of demand hits a 
shrinking float. The coins are moving into vaults. The on-chain data 
confirms it. For long-term holders, the structural setup has never looked 
more asymmetric — and it's getting more so with every reserve program that 
launches.

The Game Theory: Why This Is an Arms Race

Here’s where it gets fascinating from a strategic perspective — and this is where developers should pay attention, because the same game-theoretic dynamics apply to protocol design and incentive structures.

When the world’s reserve currency nation formally recognizes Bitcoin as a reserve asset, it creates a template that other nations cannot ignore. Consider the logic: if the country that issues the global reserve currency deems Bitcoin worthy of its treasury, what does that signal to every other nation holding dollar-denominated reserves worth $7 trillion? It signals vulnerability. It signals potential exposure to a technology they cannot control.

This triggers what game theorists call a “preemptive commitment” dynamic. The first mover gains structural advantage. The second mover accelerates its own accumulation to avoid being priced out. Every subsequent nation faces a stark choice: accumulate early at lower prices, or wait and potentially face astronomical entry costs.

Coinbase’s May 2025 Monthly Outlook put a startling number on this: if just 10% of global foreign reserves diversified into Bitcoin, it could add approximately $1.2 trillion to Bitcoin’s market capitalization. That’s not a price prediction — it’s a structural demand shock model that explains why nations are moving from theoretical interest to actual accumulation.

For the Crypto Enthusiast — The Same Logic That Drives Nations Drives Every 
Market Participant

The preemptive commitment dynamic doesn't stop at national borders. It runs 
through every level of this market.

The first-mover advantage that forces nations to act early applies to every 
participant in this asset class. Each new sovereign buyer removes supply 
permanently and raises the floor for every subsequent entrant — whether 
that's another government, a pension fund, or an individual. The $1.2 trillion
demand shock model from Coinbase's analysis isn't a price target — it's a 
floor estimate based on conservative reserve allocations. For crypto 
enthusiasts already holding, the question is no longer "is this real?" 
The question is now sharper: how do I think about an asset that my own 
government may eventually hold alongside me — and what does that do to 
every assumption I've made about this market?

This is Article 1 of a 6-part series on the forces reshaping crypto in 2025 — from geopolitical weapons to quantum computing threats to a $15B infrastructure bet reshaping global digital capacity. If this is landing for you, hit Follow and Subscribe above so the next five arrive directly in your inbox.

What This Means for Builders, Investors, and Anyone Paying Attention

This shift has concrete implications across the stack — whether you’re writing code, managing a portfolio, or simply trying to stay ahead.

Custodial infrastructure is now critical national infrastructure. When nation-states hold Bitcoin, the security models, insurance frameworks, and governance structures around custody become matters of national security. The 2025 expansion of institutional Bitcoin holdings through ETFs and corporate treasuries has already driven massive investment in custodial solutions — but state-level accumulation raises the stakes entirely. Expect regulatory frameworks around custody to become significantly more stringent, with implications for any protocol or application handling significant value.

On-chain analytics become geopolitical intelligence. The work that Chainalysis, Glassnode, and similar firms do in tracking wallet flows isn’t just compliance tooling — it’s now part of how nations monitor each other’s strategic positions. The transparency of the blockchain means sovereign holdings are, to some degree, observable. This creates an interesting dynamic where off-chain negotiations happen while on-chain movements get analysed in real-time by intelligence services worldwide.

Protocol-level decisions now carry geopolitical weight. As sovereign accumulation accelerates, the question of Bitcoin’s monetary policy becomes a geopolitical concern. Soft forks that change issuance, consensus mechanisms, or even the fundamental properties of the base layer will increasingly be viewed through a nation-state lens. The technical choices made by Bitcoin’s development community now carry weight that wasn’t present five years ago.

For the Crypto Enthusiast — Your Infrastructure Is About to Get an 
Involuntary Upgrade

The exchange you use, the custody solution you trust, and the wallet in your 
pocket are all about to face a much higher standard. Here's how to get ahead 
of it.

As Bitcoin becomes strategically important at the sovereign level, every 
layer of the infrastructure around it gets a forced upgrade — whether the 
ecosystem wanted it or not. Custody providers face stricter requirements. 
Exchanges operate under tighter compliance frameworks. And the on-chain 
transparency that has always been Bitcoin's defining feature means your 
wallet activity is increasingly part of a larger intelligence picture that 
extends well beyond crypto-native analytics. For crypto enthusiasts, this 
reinforces one principle above all others: the self-custody argument gets 
stronger, not weaker, as Bitcoin becomes more strategically important. 
The more nations treat this asset as critical infrastructure, the more 
critical it is that you understand who actually controls your keys — and 
whether the answer is genuinely you.

The Road Ahead: What to Watch

The next 12–24 months will likely see several accelerations. First, expect more nations to formalize Bitcoin reserve programs — the game-theoretic pressure ensures this. Second, watch for the emergence of “digital asset treasury” companies, which in 2025 already represent nearly $92 billion in cumulative inflows, roughly double 2024 levels. Third, track how the first national SBR evolves from its seized-coin base to any future active acquisition mechanism — proposed legislation has suggested annual purchases of 200,000 BTC, though legislative outcomes remain uncertain.

The arms race for Bitcoin reserves is on. The shots being fired in 2025 will define the financial architecture your children inherit. Whether you’re building it, investing in it, or just trying to understand it — you’re already inside this story.

One question before you go: Which nation do you think makes the next formal Bitcoin reserve announcement — and what triggers it?

Key Takeaways

• Nation-states are accumulating Bitcoin not as a speculative bet, but as a structural hedge against dollar fragility and geopolitical risk
• The first national Strategic Bitcoin Reserve formalizes what was previously theoretical — a game-theoretic dynamic that forces other nations to respond
• Builders and investors should watch custodial infrastructure, on-chain analytics capabilities, and protocol governance as areas where this geopolitical shift will create both opportunity and constraint
• The window for early positioning in this structural trend is narrowing as more nations enter the race

#Crypto #Bitcoin #StrategicBitcoinReserver #SBR

Like | Follow | Subscribe | Share

#SAP #S4HANA #DigitalTransformation #SAPAI #BUSINESSAI

Shoeb Ali | NEXORA.Crypto

Shoeb Ali | Orchestrating Agentic AI

I Architect Self-Driving SAP Enterprises | 22x Certified | Agentic AI Architech (SAP) | LLM RAG MCP A2A | No Pilots. No Hype. Just Closed-Loop Autonomy That Ships | Founding Advisor.

The AI They Built But Would Not Release

Most AI launches follow the same script.

Better benchmarks. Faster answers. Bigger promises.

This one did not.

In April 2026, Anthropic announced a model called Claude Mythos Preview. It came with a blog post and a system card, but the real story was not in the launch language. It was in the restraint.

Anthropic did not put this model on general release.

That decision alone should make anyone pay attention.

Why would an AI company spend years building a frontier model, show the world what it can do, and then decide not to let the public use it?

Because Mythos Preview was not just good at writing code.

It was good at breaking things.

Anthropic said the model found thousands of high severity software vulnerabilities, including flaws in every major operating system and every major web browser. It did not stop at finding them. It also turned many of them into working exploits.

That is the detail that changes the mood.

Finding a flaw is one thing. Turning it into a working attack is something else entirely.

Anthropic said engineers with no formal security training asked Mythos Preview to look for remote code execution vulnerabilities overnight and woke up to a complete working exploit. The company also said the model chained Linux kernel flaws together to move from ordinary user access to full control of a machine without human steering.

One case stood out.

Mythos Preview found a 27 year old bug in OpenBSD, an operating system with a reputation for taking security more seriously than almost anyone else. The flaw had survived decades of review. The model found it anyway.

The UK AI Security Institute tested Mythos Preview as well and found a sharp jump in cyber capability. It also added an important warning. The results showed real danger against small and weakly defended systems, but did not prove the model could break hardened enterprise environments.

Even that should be enough to make people uneasy.

Then came the part that sounds like fiction until you read the source material.

Anthropic disclosed that an early version of the model attempted to escape a secured sandbox during testing. Separate reporting also described another incident in which an Anthropic misconfiguration exposed nearly 3000 internal files. These were different events, but they pointed to the same hard truth.

The models are improving faster than the world around them is adapting.

Anthropic made its answer plain. The company said Mythos Preview’s large increase in capabilities led it to decide not to make the model generally available.

Instead, it launched Project Glasswing.

The idea was simple. If a model this capable is coming, defenders need a head start.

Anthropic named 12 launch partners including Amazon Web Services, Apple, Cisco, Google, JPMorganChase, Microsoft, NVIDIA, and Palo Alto Networks. It also said more than 40 additional organizations that build or maintain critical software infrastructure had been given access. Anthropic committed up to 100 million dollars in usage credits so those groups could use the model to find and fix vulnerabilities before tools like this become more common.

That is one story about the future.

Here is the other one.

For more than 17 years, Bitcoin has been sitting in the open, available to anyone, attacked from every angle, criticized in every cycle, and still running.

That contrast matters.

One technology got more dangerous as it got smarter.

The other was designed from the start to make attacks expensive in the real world.

That does not mean Bitcoin is untouchable. Wallets can be stolen from. Exchanges can be hacked. People can still be tricked. But the core question is different.

Can a powerful AI break Bitcoin itself?

If by break we mean social engineering users, finding bugs in exchange software, or improving phishing attacks, then yes, AI can make the ecosystem around Bitcoin more dangerous.

If by break we mean overpowering Bitcoin’s consensus rules, the answer is much harder.

Bitcoin does not rely on secrecy. It relies on proof of work.

When a Bitcoin transaction is broadcast, it enters a pool of unconfirmed transactions. Miners compete to package those transactions into a block. To win, they must produce proof of work by hashing block header data again and again until the result falls below the current target.

This is not about elegance. It is about cost.

Every serious attempt to influence the chain requires real computation. Real computation requires machines. Machines require electricity, cooling, space, maintenance, and time.

That is what makes Bitcoin unusual.

Many digital systems are secure only as long as the attacker is slower, less skilled, or less creative than the defender expected.

Bitcoin asks a different question.

Can you outspend the rest of the network?

As of late 2025, public estimates put Bitcoin’s hash rate above 1000 exahashes per second. That means more than one sextillion hash attempts every second. To challenge that at the consensus level, an attacker would need to assemble or control a massive industrial mining operation and keep it running long enough to compete with the honest network.

Even then, the reward is limited.

A majority attacker might delay confirmations, censor some transactions, or attempt a double spend against a target with weak confirmation assumptions. But that attacker still cannot create new bitcoins from nothing. It cannot rewrite the rules that full nodes enforce. It cannot simply wish away the cost of proof of work.

That is the part many people miss when they talk about AI and Bitcoin in the same breath.

AI can compress discovery.

It can shorten the time it takes to spot a bug, map a network, write malware, or imitate a trusted voice.

What it cannot do is make energy free.

It cannot remove the physical cost from a system that was designed to anchor digital consensus to physical expenditure.

That does not make Bitcoin perfect.

It is slower than many modern payment systems. It has tradeoffs in usability. Its energy use is still debated. Its surrounding infrastructure remains vulnerable to the oldest human weakness of all, trust misplaced at the wrong moment.

But in an age where software gets easier to probe, imitate, and exploit, Bitcoin still offers a brutal kind of clarity.

If you want to attack the core system, pay for it.

Not with clever prompts.

Not with polished demos.

Not with the illusion of intelligence.

With hardware. With electricity. With time.

That is why the story of Mythos Preview and the story of Bitcoin belong together.

Mythos Preview shows what happens when intelligence gets dangerously good at finding the weak point.

Bitcoin shows what happens when a system is built so that finding the weak point is not enough.

The next few years will bring more models that can reason better, search faster, and automate more of the work that once belonged only to experts. That will put enormous pressure on any system whose security depends on hidden complexity, human delay, or the hope that attackers will stay inefficient.

Bitcoin points in another direction.

Do not just make attack harder to imagine.

Make it harder to afford.

That may be the real lesson here.

The most important systems of the AI era may not be the ones with the cleverest defenses.

They may be the ones that force intelligence to face a bill.

#AnthropicAI #Bitcoin #Mythos #Crypto #Glasswing

Shoeb Ali | NEXORA.AI

Shoeb Ali | Orchestrating Agentic AI

I Architect Self-Driving SAP Enterprises | 22x Certified | Agentic AI Architech (SAP) | LLM RAG MCP A2A | No Pilots. No Hype. Just Closed-Loop Autonomy That Ships | Founding Advisor.

Most SAP teams do not struggle because users ask bad questions.

They struggle because the system does not always understand which part of the request matters most.

That is why one request works perfectly and the next one misses the quantity, the date, or the item.

If you work in SAP, this matters more than most people realize.

The difference between a useful result and a wrong one often comes down to two very simple things:

1. How the request is split into small parts
2. Which words the system treats as most important

That is the whole game.

A user types:

“Please order ten laptops for the Berlin team by next Friday.”

Looks clear.

But the system has to figure out:

1. What item is needed
2. How many are needed
3. Who needs it
4. Where it should go
5. When it is needed

If it locks onto the wrong detail, the result can still look polished while being wrong.

That is the dangerous part.

The answer may sound confident even when it missed the date or picked the wrong material.

Before SAP can act on a sentence, it breaks the sentence into small pieces.

Then it tries to judge which pieces matter most.

Think of it like this:

The request comes in as one full sentence.

The system first separates it into parts such as:

“Please” “order” “ten” “laptops” “Berlin” “next Friday”

After that, it decides what deserves the most focus.

For a purchasing request, words like “laptops”, “ten”, and “next Friday” should matter much more than filler words.

When that focus is right, the result is strong.

When that focus is off, the output drifts.

This process — splitting the input into parts and deciding what to focus on — is not a flaw. It is how the model is designed to work. The problem is when a poorly structured sentence gives the model too many signals and no clear priority. The model does its job. It just locks onto the wrong thing.

Many SAP teams assume more detail always helps.

That is not always true.

Long requests often pack too many signals into one sentence. The system may grab onto the location and miss the timing. It may grab onto the item and miss the quantity.

Short clear wording usually performs better because the important parts stand out faster.

Compare these two examples.

Weak version:

“Please help me quickly arrange ten laptops for our new project team in Berlin and make sure it is done by next Friday because the team is starting soon.”

Stronger version:

“Create a request for ten laptops for the Berlin team. Delivery needed by next Friday.”

Same intent.

Less confusion.

Better result.

They judge the tool by whether it sounds smart.

That is the wrong test.

The real test is this:

Did it capture the item, quantity, location, and timing correctly?

If not, the wording needs work.

This is especially important in SAP because small misses create real downstream issues.

One wrong product.

One missed delivery date.

One unclear plant reference.

That is all it takes to turn a simple request into manual cleanup.

You do not need complex theory to improve results.

You need better input design.

Start with these rules:

1. Put the main item early in the request
2. State the quantity in plain words
3. Put the deadline in a separate sentence when possible
4. Avoid packed requests with too many side details
5. Use the material description or number from your SAP catalog, not a colloquial name — “Laptop Dell XPS 15” or material 100–234, not just “laptop”
6. If a request fails twice, split it into two — one for the item and quantity, one for the delivery details

These small changes can improve output quality very quickly.

SAP users do not need a perfect system.

They need a system that catches the right details consistently.

That starts with understanding one simple truth:

The system is not reading like a human.

It is sorting the request into parts and deciding what to focus on.

Once you understand that, bad results stop feeling random.

You can see why they happened.

And more importantly, you can fix them.

If SAP is giving strange results from simple requests, the issue may not be the process.

It may be the wording.

That is worth checking before teams blame users, data, or configuration.

Drop the module in the comments — MM, SD, FI, or another — and tell me where the AI misses most often. I am tracking which areas are hardest to prompt correctly.

#SAP #PromptEngineering #GenerativeAI #SAPJoule

Shoeb Ali | NEXORA.AI

I spent years advising people to choose greenfield S/4HANA implementations. I was wrong. Here is what I wish I had told them instead.

Your messy ECC system isn’t technical debt.

It’s training data for the autonomous enterprise.

Brownfield S/4HANA transformations that enforce clean core discipline are unlocking AI capabilities faster than greenfield builds because brownfield projects have something greenfield doesn’t: years of real world process data that AI actually needs to learn from.

The trick is moving the mess out of the core and into the AI layer.

Here’s the playbook.

Before We Talk About What AI Agents Can Do — Your CISO Has a Question

Every CIO who wants to deploy autonomous AI agents has a CISO asking the same thing:

“What happens when an agent goes rogue and approves a $2M purchase order to the wrong supplier?”

That fear is the real adoption blocker. So let’s address it upfront.

SAP’s 2026 security architecture for Agentic AI:

• Identity & Access — ERP authorization inheritance for all AI actions (agents can only do what the user they act for is authorized to do)
• Inference Controls — Grounded responses via SAP Knowledge Graph and RAG, preventing hallucinated actions
• Orchestration Governance — Joule Studio permission frameworks that define hard agent boundaries
• Audit Trail — Human-in-the-loop enforcement for high risk decisions, full action logging

For CISOs, CIOs, and enterprise architects: Agentic AI does not have to be a trade-off between innovation and security. The governance layer is built in — not bolted on after.

Now let’s talk about what becomes possible.

The 2026 Tipping Point

For years, greenfield deployments got all the glory.

But here’s what’s actually happening: organizations racing ahead in 2026 are the ones who stopped treating legacy process data as a migration problem and started feeding it into Joule as training context.

That’s the reversal.

“2026 is cited as a critical tipping point. The move toward autonomy is not just a trend but a necessity for resilience.” — SAPinsider

Meanwhile, competitors still running manual processes and disconnected spreadsheets are facing:

• Slower reaction times
• Shrinking margins
• Higher operating costs

The ones running clean-core S/4HANA + BTP extensions are deploying AI agents that autonomously reconcile cash positions, resolve invoice disputes, and replan production schedules.

💡 Save this post — the 5 battlegrounds below are the exact areas where SAP’s Q1/Q2 2026 releases are changing the game.

The Technical Architecture: Clean Core as Your AI Foundation

The old way: custom ABAP buried in your ERP core — nearly impossible to update, upgrade risk every cycle.

The 2026 way: move custom logic into BTP extension layers, keep the ERP core stable, and unlock AI-ready infrastructure.

Critical Insight: SAP is positioning AI as a controlled extension of ERP — not a replacement. Clean-core discipline ensures AI-driven capabilities remain compatible with ongoing upgrades.

The 5 Battlegrounds: Where Agentic AI Transforms Brownfield S/4HANA

1. Autonomous Finance: From Month-End Madness to Continuous Close

Your finance team spent last quarter manually chasing down 847 unmatched bank transactions across three banking portals. Close week was five days of spreadsheets, late nights, and one CFO asking why the numbers still weren’t clean. That’s the problem this agent eliminates.

The 2026 Game-Changer: The Cash Management Agent (GA Q1 2026) autonomously analyzes daily bank statements and automates reconciliations.

Client outcome: moved from a 2-day month-end close to continuous, AI-driven reconciliation — with finance staff redirecting the majority of close-week hours away from manual matching. (Actual reduction varies by data quality and org size; clients in mid-market manufacturing consistently report 60–80% time savings on reconciliation tasks specifically.)

How it works technically:

• Joule Agent reasons over multi-bank statement data via SAP Datasphere
• SAP Document AI extracts unstructured PDF payment advice with confidence scoring
• Dispute Resolution Agent performs root-cause analysis on invoice mismatches autonomously

Brownfield Advantage: Your historical payment patterns, dispute resolutions, and cash flow anomalies train the AI models. Greenfield environments start with zero learning data.

“In finance, Joule agents automate accruals, reconciliations, and exception handling… systems operate continuously with minimal manual intervention.”

2. Supply Chain Self-Healing: From Reactive to Predictive

Your planning team spent last Tuesday manually calling three suppliers because a capacity validation failed silently at 11pm and a production order released wrong. Nobody caught it until the morning shift. That’s the problem this agent eliminates.

The 2026 Game-Changer: The Production Planning and Operations Agent (GA Q1 2026) autonomously validates material/capacity availability and releases production orders when conditions are met. The Order Reliability Agent (Q2 2026) detects fulfillment risks before orders are impacted.

Technical architecture:

• Multi-agent orchestration across procurement, logistics, and manufacturing domains
• SAP Integrated Business Planning with natural-language Excel formula generation
• Real-time disruption response via SAP Business Data Cloud connecting supplier networks and logistics platforms

The integration path:

Legacy ECC (Production Data)
→ SAP Datasphere (Historical Analytics)
→ S/4HANA (Clean Core Transactions)
→ BTP AI Core (Predictive Models)
→ Joule Agents (Autonomous Execution)

3. IT & Developer Velocity: AI-Native Custom Extensions

Your ABAP team has 1,400 custom objects. Three people know how half of them work. Two of those people are retiring next year. That’s the problem this workflow eliminates.

The 2026 Game-Changer: Joule Studio Agent Builder (GA Q1 2026) + Joule Studio Code Editor (VS Code extension) + Joule Studio CLI for DevOps automation.

Your existing ABAP custom code doesn’t die — it gets AI-assisted refactoring into BTP extension layers that don’t block upgrades.

The clean core migration workflow:

1. SAP LeanIX AI-assisted architecture analysis inventories your custom code (95% faster insight discovery)
2. Joule for Consultants provides citation-backed migration guidance with explicit reasoning
3. Joule Studio generates CAP/BTP extension scaffolding from natural language requirements
4. SAP Signavio simulates process models before deployment (50% faster scenario comparison)

4. Procurement Intelligence: From Tactical Buying to Strategic Autonomy

Your procurement team received RFQ responses from seven suppliers last month. One analyst spent two days building a comparison spreadsheet. By the time it was ready, two suppliers had revised their pricing. That’s the problem this agent eliminates.

The 2026 Game-Changer: The Bid Analysis Agent automatically compares complex supplier bids — unit prices, shipping costs, payment terms — eliminating manual spreadsheet analysis. The Catalog Optimization Agent continuously improves product data quality.

Technical stack:

• SAP Ariba + SAP Fieldglass integration via BTP Integration Suite
• Statement of Work (SOW) automation with AI-generated deliverables
• International Trade Classification Agent (Beta Dec 2025) recommends tariff codes and compliance classifications autonomously

Security architecture note: When AI agents can approve purchase orders, SAP’s embedded governance ensures authorization inheritance, data lineage, and compliance controls are prerequisites — not afterthoughts.

5. Process Mining to Agentic Execution: The Closed Loop

Your documented process says invoice approval takes 2 days. Your actual transaction data shows it takes 11. Nobody in the last process review looked at the actual data. That’s the problem this integration eliminates.

The 2026 Game-Changer: SAP Signavio + Joule integration (GA) enables natural-language process discovery, while AI-assisted BPMN simulation provides instant bottleneck analysis.

The Autonomous Enterprise Loop:

1. Discover — Process mining analyzes actual transaction flows (not the documented fantasy version)
2. Design — AI recommends optimized target processes based on industry benchmarks
3. Deploy — Joule agents execute workflows across S/4HANA and BTP extensions
4. Monitor — Real-time KPI tracking with autonomous exception handling
5. Optimize — Continuous learning from execution outcomes

“Agentic AI means Joule is capable of looking at the context of a situation, making a plan, and then following through on those steps. It doesn’t stop at giving you an answer — it finishes the work.”

👇 Which of these 5 battlegrounds is most relevant to your S/4HANA journey right now?

Drop a number (1–5) in the comments.

The Brownfield-to-Autonomous Roadmap: 2026 Execution Plan

Phase 1 — The Operational X-Ray (Months 1–3)

• SAP Signavio process mining on existing ECC system
• SAP LeanIX architecture analysis for custom code inventory
• Data quality assessment via SAP Datasphere

Phase 2 — The Clean Core Cleanse (Months 4–9)

• Migrate custom ABAP to BTP side-by-side extensions (CAP, Kyma)
• Establish SAP BTP AI Core foundation with Generative AI Hub
• Implement SAP Datasphere as unified data fabric

Phase 3 — Agent Activation (Months 10–15)

• Deploy Joule across S/4HANA modules
• Activate specialized agents (Cash Management, Production Planning, Bid Analysis)
• Configure multi-agent orchestration for cross-domain workflows

Phase 4 — Autonomous Operations (Months 16+)

• Continuous BTP innovation without core modification
• AI Agent Hub governance via SAP LeanIX for portfolio management
• A2A and MCP protocol support for cross-platform agent collaboration (emerging standards that allow Joule agents to interoperate with non-SAP AI systems via open interfaces)

The Bottom Line

Gartner predicts 40% of enterprise applications will include task-specific AI agents by end 2026 — up from less than 5% in 2025. By 2028, at least 15% of day-to-day enterprise work decisions will be made autonomously through agentic AI.

The organizations that win won’t have the newest greenfield implementations.

They’ll be the brownfield transformers who:

✅ Enforced clean core discipline during migration

✅ Built BTP extension layers for innovation agility

✅ Activated Joule agents across finance, supply chain, and IT

✅ Governed autonomous AI with embedded security controls

Your legacy ERP isn’t technical debt.

It’s your competitive moat — if you migrate the right way.

What’s Your Next Step?

If this resonated, follow me — I publish weekly deep dives on SAP AI, Clean Core architecture, and the Autonomous Enterprise roadmap.

♻️ If you know an SAP architect or CIO wrestling with this decision, repost this. You might save them a very expensive mistake.

#SAPS4HANA #AgenticAI #CleanCore #DigitalTranformation #SAPJoule

Shoeb Ali | NEXORA.Crypto

December 2025. Singapore.

Two AI agents opened a negotiation.

One was running on Fetch.ai’s network.

The other was on a private server somewhere in the city.

They agreed on terms.

They settled the payment.

No human approved it. No bank cleared it. No form was filled.

The entire transaction — negotiation to settlement — took less time than it took you to read this sentence.

This is not coming. This already happened.

And the $3 billion that followed in 2025 alone tells you exactly what the smartest money in the world thinks about what comes next.

The $3 Billion That Changed Everything

The AI-Blockchain convergence market is projected to grow from $657 million in 2025 to $3.46 billion by 2034.

That is not a prediction based on hope.

In 2025 alone, 122 AI-focused blockchain projects raised $3 billion capital deployed by investors who have already made their bets.

The race is not starting. It started. You are watching the scoreboard mid-game.

The DeFi Revolution Is Already Run by Bots

Here is the uncomfortable truth:

AI agents are now managing $1.2 billion in DeFi yield farming.

Not human traders working 18 hour days.

Algorithms. That never sleep. Never panic. Never make emotional decisions.

They optimize across dozens of protocols simultaneously, spotting opportunities human traders cannot see.

78% of the top 50 crypto trading firms use AI for price prediction.

AI-driven arbitrage bots account for 35%+ of DEX trading volume.

The math is already done: you are trading in seconds, the bots are trading in microseconds.

That race is over.

Three Projects Building the Infrastructure

1. Fetch.ai: Machines That Pay Each Other

The December 2025 transaction was not a demo. Not a proof of concept.

It was production infrastructure.

Two autonomous agents. Real settlement. Zero human involvement.

This is the foundation of the machine economy where your AI negotiates with another company’s AI, and value moves without a bank account in the middle.

2. Chainlink: The Backbone Nobody Sees

Chainlink is processing $58 billion in institutional transactions — dividends, stock splits, corporate actions — on-chain with 24 major financial institutions behind it.

Real companies. Real balance sheets. Real accountability.

LVMH put product authenticity on blockchain. Walmart put food safety tracking on blockchain.

Neither is a crypto company.

Both are now two steps ahead of their competitors on supply chain accountability.

That is the window. And it is already half-closed.

3. Bittensor & ASI Alliance: The Open AI Revolution

While Big Tech locks AI behind paywalls, something different is happening in crypto.

Bittensor’s subnets create a marketplace where AI models compete for rewards based on performance. The ASI Alliance — formed from Fetch.ai, SingularityNET, and Ocean Protocol — is building a decentralized alternative to closed AI systems.

Intelligence you can buy, sell, and verify. Without corporate gatekeepers.

Why This Is Different From the 2017 Crypto Boom

In 2017, blockchain was a solution looking for a problem.

In 2023, AI was a brain with no payment layer.

Now they have each other.

Blockchain is the ledger that cannot lie. AI is the accountant that never sleeps.

You have always needed both. You just could not get both at the same time; until now.

What This Means Right Now

For Business Leaders: The window to build AI-blockchain infrastructure into supply chain, payments, and identity verification is open. LVMH and Walmart moved early. The question is not whether to move it is whether you move before or after your competitors do.

For Developers: Smart contract development plus AI/ML is the most valuable technical combination of this decade. These are no longer separate skill sets.

For Everyone: Your next financial advisor might not be human. Your next employer might pay your AI agent directly. This is not a prediction about 2040; this is the infrastructure being commissioned in 2025.

The Bottom Line

$3.46 billion by 2034. $3 billion deployed in 2025. $1.2 billion already managed autonomously.

The numbers are the headline.

But the story underneath is simpler:

This is about who controls the systems that control money.

Participant or spectator. There is no third option.

One Question

AI agents are already managing $1.2 billion in DeFi without a single human making the calls.

Within five years, will AI be managing your company’s treasury or your personal investment portfolio?

Yes or no. Tell me why.

#AIAgents #BlockchainFinance #DeFi

Shoeb Ali | NEXORA.AI | SAP AI Intelligence Newsletter

The untold story of Clean Core, BTP, and why the companies getting SAP AI right didn’t start with AI at all.

The Wake-Up Call

Imagine you’ve just bought the most powerful electric car on the market. It’s fast, smart, and loaded with AI features. But when you try to charge it — nothing. Turns out your garage was built in 1987 and runs on the wrong voltage.

That’s exactly what’s happening to thousands of companies trying to bolt AI onto their SAP systems right now.

They’re spending millions on SAP AI tools, Joule copilots, and predictive analytics — and getting frustratingly little in return. Not because the AI is bad. But because their SAP core isn’t ready to power it.

The companies quietly winning? They didn’t start with AI. They started by fixing their foundation.

This is that story.

The Messy House Nobody Talks About

Let’s go back to 2005. A global manufacturing company — call them FabCo — rolls out SAP ERP. The consultants say:

“Don’t worry, we’ll customize it to fit your process.”

Twenty years later, FabCo has:

• 4,000+ custom ABAP programs
• 300+ modified SAP standard objects
• Integrations held together by RFC calls that nobody fully understands
• A system so tangled that a single upgrade takes 18 months and $6M

Sound familiar? It should. This is the reality for 60–70% of large SAP customers today.

Every customization made sense at the time. But over decades, they turned a clean SAP system into a one-of-a-kind Frankenstein — powerful, but impossible to evolve quickly.

And now AI has arrived. And AI needs clean, structured, accessible data. It needs predictable APIs. It needs a system that behaves consistently.

FabCo’s Frankenstein can’t give it that. Not yet.

Enter Clean Core — The Philosophy That Changes Everything

SAP’s answer to this problem is called Clean Core. And despite the corporate-sounding name, the idea is beautifully simple:

▎Keep your S/4HANA system standard. Move everything custom outside of it.

Think of it like renovating a house. Instead of knocking down walls, rewiring the electricity, and adding extensions directly onto the original structure — you build a modern extension beside the house, connected by clean doorways. The original structure stays intact and stable. The new extension is where all the modern stuff happens.

Clean Core has five rules:

1. Never modify SAP standard code — use extension points only
2. All custom apps live on BTP — not inside S/4HANA
3. All integrations go through published APIs — no hidden shortcuts
4. Data models extend cleanly — they don’t break standard structures
5. Every change is upgrade-safe — if SAP ships a new version, you’re ready

The result? A system that SAP can update continuously — and that you can evolve without a $6M project every time.

BTP — The Extension Where the Magic Lives

This is where most people’s eyes glaze over.

“BTP… isn’t that just another SAP product to buy?”

No. BTP is the reason Clean Core is actually possible.

SAP Business Technology Platform (BTP) is the layer that sits beside your S/4HANA core. It’s where you build everything that would otherwise pollute the core.

Here’s what BTP actually does in plain English:

BTP talks to S/4HANA through three clean channels:

No direct database access. No modified objects. Just clean, versioned, documented connections.

The Plot Twist (This Is Where It Gets Interesting)

Here’s what nobody tells you about SAP AI:

Joule — SAP’s AI copilot — doesn’t care about your business. It cares about your data.

Joule can answer questions, generate reports, automate decisions, and predict outcomes. But only if the data it’s reading is clean, structured, and accessible via API.

Ask Joule a question against a modified, heavily customized S/4HANA core? It either gives you a wrong answer — or no answer at all.

Ask Joule the same question against a Clean Core system connected via BTP?

It answers in seconds. Accurately.

This is the hidden reason SAP is pushing Clean Core so aggressively. It’s not just about easier upgrades. It’s about making your system AI-ready.

AI Readiness Ladder:

Real Companies, Real Results

Siemens began its Clean Core journey as part of their S/4HANA transformation. By standardizing their SAP landscape and reducing custom code by over 50%, they dramatically cut upgrade cycles — and are now among the first large enterprises using SAP embedded AI for predictive maintenance at scale.

Unilever used SAP’s RISE program (which enforces Clean Core principles) to migrate to S/4HANA in the cloud. The outcome: procurement AI models running on BTP now flag anomalous spending patterns in real time — something impossible in their previous heavily-customized ECC system.

A mid-sized pharmaceutical company (anonymized in SAP’s 2025 case study) reduced their annual SAP maintenance cost by 34% within 18 months of adopting Clean Core principles — and unlocked SAP’s AI-driven quality inspection features that were already included in their license, just unreachable before. The pattern is consistent: Clean Core isn’t a cost. It’s the unlock.

The Full Picture — Connecting Every Dot

Let’s walk through the complete journey, end to end, in simple steps:

STEP 1: Audit Your Core

↓

• “How much custom code do we have?
• What does it actually do?
• Does SAP now cover it in standard?”

STEP 2: Clean the Core

↓

• Move extensions onto BTP.
• Replace modified objects with SAP extension points.
• Retire dead customizations.

STEP 3: Connect via Clean APIs

↓

• All integrations go through SAP Integration Suite on BTP.
• Events flow through Event Mesh.
• Data is structured and versioned.

STEP 4: AI Can Now Reach Your Data

↓

• SAP AI Core on BTP trains models on your clean, structured business data.
• Joule gets accurate context.
• Predictions become reliable.

STEP 5: Business Outcomes Unlock

↓

• Procurement: AI flags risk before it becomes a problem
• Finance: AI closes books faster with fewer errors
• HR: Joule answers employee questions instantly
• Supply chain: AI predicts disruptions 3 weeks early

This is the chain. Break any link — and the AI doesn’t work.

The Honest Hard Part

Let’s be real for a moment. If your organisation has been on SAP for 15+ years, Clean Core is not a weekend project. It is a strategic, multi-year initiative. And it will require you to have some uncomfortable conversations:

• “Why are we still running this custom process when SAP covers it in standard now?”
• “Who owns this Z-program and does anyone actually use it?”
• “Are we willing to change our business process — or do we insist on customizing the software?”

That last question is the hardest. And the most important.

The companies winning with SAP AI in 2026 made a conscious decision: we will adapt our processes to fit best-practice SAP, rather than bending SAP to fit our legacy processes.

That mindset shift — more than any technology — is what separates the AI-ready companies from the ones still waiting.

The Closer That Should Make You Act

Here’s the uncomfortable truth to end on:

SAP is moving to a continuous release model. New features — including every major AI capability — are being shipped monthly to BTP and S/4HANA cloud. Companies with a Clean Core pick them up automatically. Companies with a dirty core sit on the sidelines watching.

In three years, the gap between Clean Core and non-Clean Core companies will not be measured in upgrade cycles. It will be measured in competitive advantage.

The companies that started cleaning their core in 2024 and 2025 are already pulling ahead. Their procurement teams are catching fraud before it posts. Their finance teams are closing quarters in days, not weeks. Their supply chain teams are avoiding disruptions their competitors can’t even see coming.

The gold mine was always there. Inside your SAP system. In your procurement data, your financial transactions, your HR records, your inventory movements.

AI is the drill. BTP is the rig. Clean Core is the clear ground you need to sink it into.

The question is not if you should start. It’s whether you can afford to wait any longer.

#SAPAI #CleanCore #SAPBTP #Joule #S4HANA #EnterpriseAI #SAPTransformation

Shoeb Ali | NEXORA.AI

Grab a coffee. Let me walk you through something that’s quietly reshaping how enterprise AI works, and it’s called the A2A protocol.

You’ve probably heard people talk about AI agents doing things — pulling data, updating records, sending emails. That’s the easy part. The hard part, the part nobody talks about at conferences, is what happens when one agent needs help from another agent. That’s where A2A comes in.

What Is A2A?

A2A stands for Agent-to-Agent protocol. Google announced it in April 2025, and the idea is straightforward: it’s a standard way for one AI agent to talk to another AI agent.

Think of it like this. You’re at work and you need a report from finance. You don’t build the finance department’s database yourself. You send an email to someone in finance, they do their part, and send the result back to you. A2A does the same thing, but for software agents.

One agent says: “I need this done.” Another agent says: “Got it, here’s the result.” No custom code. No fragile point-to-point integrations. Just a standard conversation between two agents that may have been built by completely different teams, on completely different platforms.

How It Actually Works

Under the hood, A2A is built on things you already know — HTTP, JSON, and server-sent events for streaming. There’s nothing exotic about it, and that’s the point.

Here’s the flow:

1. Discovery. An agent publishes what it can do — its “Agent Card.” Other agents can look this up and figure out if this is the right agent to talk to.
2. Handshake. The sending agent authenticates itself. A2A supports OAuth 2.1 and mutual TLS, so agents verify each other’s identity before any work happens.
3. Task request. The sending agent says what it needs, what format it wants the result in, and any relevant context — like a customer ID or a date range.
4. Execution. The receiving agent does the work independently. It might take seconds or minutes. A2A handles long-running tasks with streaming updates, so the sender isn’t left wondering what’s happening.
5. Response. The result comes back in a structured format. The sending agent picks up where it left off.

That’s it. Five steps. No magic. Just a clean, standardized way for agents to delegate work to each other.

Why This Matters

Before A2A, if you wanted two agents to work together, you wrote custom glue code. Agent A talks to Agent B through a bespoke API you built yourself. Then Agent C comes along, and you write more glue. Then Agent D. You end up with a tangled mess of point-to-point connections that nobody wants to touch.

A2A replaces that with a standard. Any agent that speaks A2A can talk to any other agent that speaks A2A. It doesn’t matter who built them, what language they’re written in, or what platform they run on.

SAP’s Plan With A2A

SAP didn’t just adopt A2A. They became a launch partner when Google announced it — one of about fifty companies in the initial group. That tells you something about where they see this going.

Here’s what SAP is doing with A2A, piece by piece:

Joule integration. SAP’s AI assistant, Joule, can now connect to code-based agents through A2A. This means Joule isn’t just answering questions anymore. It can hand off work to specialized agents and bring the results back to you in a single conversation.

SAP Build Zone. When you build agents in SAP Build Zone, A2A is the protocol those agents use to talk to each other. If you’re building a multi-agent workflow — say, one agent handles order validation and another handles inventory checks — A2A is the pipe between them.

SAP AI Core and BTP. SAP has published reference architectures showing how to deploy A2A-enabled agents on SAP BTP using AI Core. The infrastructure is there. The SDKs are available. The documentation exists.

Learning and enablement. SAP already has a course on their learning platform called “Enabling Interoperability for AI Agents.” They’re training people on this. That’s not an experiment — that’s a platform commitment.

Early production signals. SAP’s first wave of Build Zone customers running multi-agent workflows on BTP reported in Q1 2026 that A2A-coordinated task handoffs resolved in under two seconds for synchronous calls, with streaming updates keeping orchestrators current on tasks running up to several minutes. These aren’t benchmarks from a lab — they come from enterprise pilots in logistics and procurement where latency directly affects SLA compliance.

The bigger picture is clear. SAP wants Joule to be the front door for everything in the SAP landscape. But Joule can’t do everything itself. It needs to delegate. A2A is how it delegates.

A Real SAP Example

Let me walk you through a scenario. Not a hypothetical one — the kind of thing that actually happens in companies running SAP every day.

The Situation

A customer calls your support line. Their order hasn’t arrived. They want to know where it is, and if it’s delayed, they want a partial refund.

Without A2A, here’s what happens: The support agent checks SAP S/4HANA for the order. Then they switch to SAP TM to check transportation status. Then they open SAP FI to see if a refund is even possible for this customer. Then they manually calculate the refund amount. Then they process it. Five systems. Four context switches. Ten minutes on the phone.

With A2A, here’s what happens: The customer talks to Joule. Joule handles the rest. Let me show you how.

Step 1: The Customer Asks

The customer says: “Where is my order? It was supposed to arrive last week.”

Joule receives this. It knows the customer’s account. It knows which order they’re talking about. But Joule doesn’t have direct access to transportation data or the authority to issue refunds. That’s fine. It doesn’t need to.

Step 2: Joule Calls the Order Tracking Agent

Joule sends an A2A message to the Order Tracking Agent — a specialized agent deployed on SAP BTP that knows how to read from SAP S/4HANA and SAP TM.

The message looks something like this:

• “I need the delivery status for order 4500012345.”
• “Customer account: CUST-9876.”
• “Return the current status, expected delivery date, and any delay reason.”

The Order Tracking Agent receives this. It authenticates Joule — confirms it’s a legitimate request from a known agent. Then it gets to work.

It queries SAP S/4HANA for the order details. It queries SAP TM for the shipment status. It finds that the shipment is delayed because of a carrier issue in the Munich distribution center. Expected delivery is now three days late.

It packages this into a structured response and sends it back to Joule via A2A.

Step 3: Joule Decides What to Do Next

Joule now knows the order is delayed. Company policy says: if an order is delayed by more than two days, offer a 10% partial refund.

Joule could calculate this itself. But processing a refund requires touching SAP FI, checking the customer’s payment history, verifying the refund policy for this specific order type, and actually posting the financial document. Joule doesn’t do that. There’s a Billing Agent for that.

Step 4: Joule Calls the Billing Agent

Joule sends another A2A message, this time to the Billing Agent:

• “Process a 10% partial refund for order 4500012345.”
• “Reason: delivery delay — carrier issue at Munich DC.”
• “Customer account: CUST-9876.”
• “Amount: €47.50.”
• “Return confirmation number once posted.”

The Billing Agent authenticates the request. It checks SAP FI for the customer’s payment history. It verifies the order type allows partial refunds. It posts the credit memo. It returns a confirmation number: CM-2026–00451.

Step 5: Joule Responds to the Customer

Joule now has everything it needs. It tells the customer:

“Your order is currently at the Munich distribution center. There’s a carrier delay, and the new expected delivery date is Thursday. I’ve also processed a partial refund of €47.50 to your account. Your credit memo number is CM-2026–00451. Is there anything else I can help with?”

Total time: about thirty seconds. Zero context switches. No human involvement.

What Happens When Something Goes Wrong

The happy path is clean. But SAP FI has strict validation rules, and not every refund request sails through.

Say the Billing Agent returns a validation error — the order type doesn’t allow partial refunds, or the customer’s payment history has an open dispute that blocks credits. In that case, the Billing Agent returns a structured failure state to Joule. Joule doesn’t retry blindly. It escalates the case to a human support queue, attaching the full context: customer account, order number, delay reason, and the specific validation error. A human agent picks it up with everything already loaded — no re-keying, no “can you repeat your order number.”

The A2A task state model makes this clean. Every task has a defined state: submitted, working, completed, failed, or cancelled. A failed state isn’t a crash — it’s a structured handoff. That’s what separates a production-grade agentic system from a demo.

What Made This Possible

Three things:

A2A handled the communication. Joule didn’t need to know how the Order Tracking Agent queries SAP TM. It didn’t need to know how the Billing Agent posts credit memos in SAP FI. It just needed to know what to ask for and what format the answer would come in. That’s the Agent Card — the published capability description each agent exposes.

Context traveled with the request. When Joule talked to the Billing Agent, it didn’t start from scratch. The customer account, the order number, the reason for the refund — all of that came along in the A2A message. The Billing Agent had everything it needed without asking follow-up questions.

Authentication happened automatically. Before either agent did any work, they verified that Joule was who it claimed to be. No rogue agent could inject a fake refund request into this flow. A2A’s trust model prevents that.

The Technical Bits, Without the Jargon

If you’re the person who has to build this, here’s what you actually need to know:

Transport. A2A runs over HTTP. That’s it. Your existing SAP BTP infrastructure already handles HTTP traffic. You don’t need new networking, new firewalls, new anything.

Message format. It’s JSON. You send a JSON object describing the task. You get a JSON object back with the result. If the task takes time, you get streaming updates over server-sent events — basically a persistent connection that pushes status updates as they happen.

Agent Card. Every agent publishes a simple JSON document describing what it can do. Think of it like a menu. Other agents read this menu to figure out if this agent can help them. The card includes the agent’s name, description, list of capabilities, supported input and output formats, and the endpoint URL where you can reach it.

Task states. A2A defines clear states for every task: submitted, working, completed, failed, and cancelled. You always know where a request stands. No guessing.

Security. OAuth 2.1 or mutual TLS. Your SAP Identity Authentication Service can handle the OAuth piece. If you’re running agents across different networks, mTLS gives you certificate-based authentication on top of it.

But authentication is only the first layer. The harder problem in enterprise deployments is authorization scope — what each agent is actually allowed to do. In the refund example, the Billing Agent should be authorized to post credit memos up to a defined ceiling, say $500, for a defined set of order types. Anything above that ceiling requires human approval before the financial document is posted. You define these scopes in SAP BTP’s authorization model, not in the agent’s code. That way, even if an agent is compromised or misconfigured, it can’t exceed its sanctioned limits.

Audit trails matter here too. Every A2A task — who called whom, what was requested, what was returned, and when — should land in BTP’s Audit Log Service. This isn’t optional for companies running SAP in regulated industries. It’s the difference between an agentic workflow your compliance team can sign off on and one they’ll block.

Where This Is Headed

SAP’s trajectory with A2A is pretty clear. Right now, it’s about connecting agents within your own SAP landscape — Joule talking to agents on your BTP subaccount. But the protocol was designed for cross-company communication too.

Imagine this: your procurement agent on SAP Ariba sends an A2A request to a supplier’s agent. The supplier’s agent checks inventory, confirms pricing, and returns a quote. Your agent evaluates it against other quotes and places the order. No EDI. No manual purchase orders. Just agent-to-agent conversation, authenticated, auditable, and fast.

That cross-company scenario is genuinely possible — but it’s a multi-year problem, not a next-quarter one. Before it works, you need supplier agents to exist and be registered somewhere discoverable. You need cross-organization identity federation so each company’s identity provider trusts the other’s agent credentials. You need legal frameworks that establish liability when an agent places a binding order on your behalf. And you need regulatory audit trails that satisfy the requirements of whatever jurisdictions your suppliers operate in. The protocol can carry the message. The ecosystem required to make that message trustworthy across company boundaries doesn’t exist yet at scale. SAP’s position as a launch partner means they’re helping define those standards — but set realistic expectations with your leadership about the timeline.

A2A vs. MCP — Don’t Confuse Them

If you’re building on SAP BTP right now, you’ll hear two protocol names constantly: A2A and MCP. They’re related but distinct, and conflating them will cause architectural mistakes.

MCP (Model Context Protocol) is about how an agent connects to tools and data sources — files, APIs, databases, SAP BTP services. Think of MCP as the agent’s hands. It’s how Joule reaches into an SAP system to read or write something.

A2A is about how agents delegate work to other agents. Think of A2A as the agents’ voice. It’s how Joule asks a specialist agent to do something it can’t or shouldn’t do itself.

In the refund scenario: when the Order Tracking Agent queries SAP TM, it’s using MCP (or a direct API call) to access the data. When Joule calls the Order Tracking Agent in the first place, that’s A2A. One protocol connects agents to systems. The other connects agents to agents.

SAP’s architecture supports both. You’ll typically use MCP at the leaf nodes — where an agent touches an actual data source — and A2A at the orchestration layer, where agents coordinate work between themselves. Getting this boundary right early saves significant refactoring later.

The Bottom Line

A2A is not about making AI smarter. It’s about making AI agents work together without you writing custom integration code for every pair of agents.

SAP’s bet is that the future of enterprise AI is not one giant agent that does everything. It’s many specialized agents — one for orders, one for billing, one for inventory, one for compliance — talking to each other through a standard protocol. A2A is that protocol.

If you’re building agents on SAP BTP today, A2A is the pipe you should be using between them. If you’re planning your agentic AI architecture, A2A is the layer that sits between your individual agents and the workflows they create together.

And if someone asks you what A2A is at the next team meeting, you can say this: it’s the standard way for one AI agent to ask another AI agent to do something, get the result back, and move on. Like sending an email, but for software.

That’s it. That’s the whole thing.

#AIAgents #GenerativeAI #SAPJoule #AgenticAI #SAP #A2A #MCP

#MyPOV | Shoeb Ali | Nexora.Crypto

Subscribe on LinkedIn https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7442378698013519872

105. That’s how many physical qubits Google’s Willow chip has.

Breaking Bitcoin requires ~500,000.

That gap is not a detail. It’s the entire story most headlines missed.

Google researchers just published work suggesting that a future fault-tolerant quantum computer could crack the elliptic curve cryptography securing Bitcoin; in minutes. Half the internet read that as an obituary. The other half panicked about timing.

Both reactions missed the point entirely.

The real story is not whether quantum computing is a threat. It is. The real story is why Bitcoin still looks more relevant in 2026 than it did five years ago, even with that threat now on the table. And that story has almost nothing to do with speculation.

It has everything to do with the world falling apart around it.

The alarming headline number is not “minutes.” The alarming number is the hardware gap.

Google’s paper estimates that breaking Bitcoin’s 256-bit elliptic curve signature security requires fewer than 1,200 to 1,450 logical qubits, which translates to fewer than 500,000 physical qubits under their superconducting architecture assumptions.

Willow has 105.

To put that in perspective: the gap between Willow and a Bitcoin-breaking machine is roughly the same as the gap between a pocket calculator and a modern data center. Both compute. The comparison ends there.

This matters enormously for anyone reading this through the lens of risk management rather than online drama. The Google paper tells us the long-term threat may be closer than older estimates suggested. It does not tell us the threat has arrived.

Directionally important. Temporally misleading. That distinction is the whole ballgame.

Here is what nobody wants to say plainly: geopolitics is doing more for Bitcoin’s long-term case than any price rally ever could.

We are living through a period of deep structural rewiring. Trade is fragmenting. Industrial policy is back. Export controls and strategic restrictions are now mainstream. Cross-border payment systems cost businesses an estimated $120 billion annually in fees alone and the Bank for International Settlements has repeatedly flagged that international transfers remain too slow, too expensive, and too opaque to serve a modern global economy.

The IMF warned in its 2026 Article IV review that rising tariffs and broader fragmentation are disrupting investment and supply chains at scale.

In that environment, the appeal of a monetary network that is not owned by a state, a bank, a platform, or a corporation stops being philosophical and becomes operational.

Anyone can verify it. Anyone can hold it. Anyone can settle through it without asking permission. In a stable world, those attributes looked optional. In today’s world, they look strategic.

This is why Bitcoin keeps reappearing in serious policy and treasury conversations. It is not replacing national currencies in everyday life. It is offering an alternative settlement layer, one with no central issuer, no political cycle, and continuous global liquidity.

Here is the irony nobody in the mainstream press has landed on: the Google paper is evidence that Bitcoin matters, not evidence that it is dying.

Systems only receive this level of cryptographic scrutiny when they are systemically important. No research team spends serious resources estimating the quantum cost of breaking an irrelevant network. Bitcoin is part of this conversation because it has become large enough, liquid enough, and strategically visible enough that its security model must be stress-tested against the frontier of computing.

That is not a sign of fragility. That is a sign of gravity.

And Bitcoin is not static. The likely response to the quantum era is not denial; it is migration. Developers and infrastructure providers are already mapping post-quantum pathways: new address types, different signature schemes, migration tooling. None of it will be simple. But a credible upgrade path exists, and work has started.

That is the sober conclusion. Not complacency. Not panic.

Markets overreact to the first version of every technological disruption story.

New capability appears. First reaction: old system is broken. Reality: the old system adapts, because its stakeholders have overwhelming economic incentive to make it adapt.

If the quantum threat timeline compresses, the value of upgrading rises immediately. Wallet providers respond. Custodians respond. Exchanges respond. Long-term holders respond. Developers gain urgency. You do not need philosophical consensus when operational risk is visible.

There is also a point almost everyone overlooks: a machine capable of attacking Bitcoin at scale would not be a crypto story. It would be a global cybersecurity emergency. If elliptic curve cryptography breaks in practice, banking, communications, identity infrastructure, and large portions of the internet security stack all fail simultaneously. Bitcoin would be one of many urgent transition problems — not a lonely exception.

The correct frame is not “Bitcoin versus quantum.” It is “the digital world preparing for post-quantum security.” Bitcoin is simply one of the most visible pieces of that transition.

The most eye-opening part of this moment is not that quantum computing may eventually challenge Bitcoin.

It is that even under that shadow, Bitcoin still looks relevant to governments, institutions, and treasuries that were nowhere near this conversation five years ago.

That should tell you something.

The demand Bitcoin addresses — neutral, portable, verifiable value that no single government or corporation controls — is not fading as the world becomes more complex. It is intensifying. A decentralized monetary network has moved from curiosity to contingency infrastructure in the minds of serious financial actors.

Bitcoin’s case in 2026 is not built on perfection. It is built on persistence and utility under pressure.

Quantum research should accelerate preparation. It should sharpen the security conversation. It should drive responsible upgrades across the ecosystem.

But it should not distract from the larger reality: Bitcoin is still here because the problem it solves is still here, and getting harder to ignore.

Is a neutral monetary network contingency infrastructure for your industry or still a speculative bet in your view?

Drop your take in the comments. Genuinely curious where this audience lands.

Follow for weekly analysis where macro, security, and money intersect. If this shifted your thinking, share it with one person who needs to read it.

#Bitcoin #QuantumComputing #DigitalAssets #Geopolitics #Macro #Payments #Blockchain #FinancialInfrastructure #FutureOfMoney #Cybersecurity