The Sovereign Intelligent Payment-Chain

March 31, 2026
|
Anil Hansjee

‍Applying AI and Data within On-Chain Payments

From the Access Era in On-Chain Payments: Leveraging the Arbitrage of Regulation…

Much like the early Internet ISP winners, the on-chain payments industry has to date focused heavily on the provision of Access. Many early projects originally utilised regulatory arbitrage to bridge fiat into digital ledgers and create an initial GTM hook. As the market and regulation has matured, this shifted toward a) tackling these same significant emerging regulatory hurdles (licensing and compliance), b) technical integrations with a variety of local fiat rails, and c) operational challenges to deliver global reliability. Historically, mainstream applications, in the US as an example, had to navigate a minefield of 50+ US state money-transmitter licenses and complex "Banking-as-a-Service" relationships. On-ramp pioneers, such as MoonPay* and Ramp Network* (both Fabric Investments,  these denoted by * from now on), became “proper” payments companies and won by building a "Regulatory Moat," making it effectively safe for others to offer a "Buy" button.
‍

Clearing this Access hurdle originally felt like an endgame for payments on-chain, but instead the industry has continued to iterate, producing several more specialised B2B technology solutions (with illustrative names from Fabric’s portfolio below):

Business models have also been merging to extend company value propositions: 

These evolutions, as important as they are, effectively remain largely a refinement of this Access phase.  Moving money on-chain currently provides such a radical "macro savings" (e.g. bypassing $50 SWIFT fees and 3-day wait times) that senders often overlook any "micro optimisations" of on-chain inefficiencies. While these immediate macro saving opportunities are large enough to mask underlying suboptimality, the industry is approaching a 2027–2028 Tipping Point that will change this.
‍

…. Towards the Intelligence Era of On-Chain Payments: Leveraging the Arbitrage of Inefficiencies

As Access providers become commoditised and consolidated (as described above), the new basis of alpha and competitive strength in on-chain payments will be routed in an intelligence layer focused on the predictive, autonomous optimisation of cost, speed, and compliance. It will address inefficiencies from an end-to-end perspective in moving funds (i.e. including treasury and back-office workflows). When a supplier saves 3% in FX fees, they rarely worry about 5 basis points of slippage. However, dollars are lost to slippage in thin liquidity, and considerable administrative drag occurs by manual reconciliations when the transaction hash fails to reconcile with their ERP system. 

This tipping point is rapidly approaching, driven by the dual tailwinds of higher volumes from B2B payments on-chain and forthcoming machine transactions. Both come with specific requirements to succeed with on-chain payments. 

The B2B payments honey pot alone is significant: While on-chain stablecoin volumes reached a staggering $35 trillion in 2025, most of this is not true end-user payments, such as paying suppliers or sending remittances. It consists mainly of trading, internal shuffling of funds, and automated blockchain activity.  McKinsey and Artemis identified $390 billion of true stablecoin payments volume in 2025 and "real-world" B2B transactions - involving the exchange of physical goods and services—accounted for approximately 60% ($226 billion) of that total. Even though this represents a massive 733% year-on-year surge,  it remains a mere "tip of the iceberg" of only about 0.014% of global B2B payment volumes of roughly $1.6 quadrillion (including intercompany treasury flows and capital markets) or only 0.12% of just 3rd party commercial B2B flows at 187$ trillion. 

Add to this opportunity the rise of agent transactions: Already by year end 2026, one-third of B2B payment workflows will utilise autonomous agents, as predicted by Forrester. Going forward, agentic payments will expand the B2B payments market by automating and replacing some current manual processes and unlocking new economic value by enabling higher transaction frequencies and new business models.  Indeed, autonomous agentic AI is well suited for B2B transactions – think supplier and buyer dispute management, invoice matching, and payment reconciliation with ERP etc. This is quite distinct from the current crop of announcements today which is focused on managing consent, authentication, authorisation, and disputes within B2C agentic flows and solved by “Human in the loop” assisted AI-agent payments (such as the recently announced Mastercard and Google Verifiable Intent initiative or the Visa Intelligent Commerce offering built on pre-authorised cards).

….The Sovereign Intelligence Era in Payments

This is a different take on sovereignty as represented in today’s geopolitical battle for hyper scaler AI under the pretence of national security. What I would like sovereign intelligence to represent within payments is the convergence of asset sovereignty (with blockchain as a tool for financial sovereignty) and execution intelligence (leveraging payments data and analytics). In the previous era within on-chain payments, we have largely used blockchain as a high-speed mirror of the legacy fiat system—a faster way to move money from point A to point B. In this era, "moving money" will become synonymous with "programming value".
‍

This era will be anchored by three pillars:

  1. Self-Custodial Control (Sovereignty): In TradFi, intelligence is at the mercy of third-party permission (whether a bank or card network) for example a better identified payment route. Instead on-chain intelligence should be sovereign. Because assets, like stablecoins, sit on a neutral, programmable ledger a business can control both the asset and the infrastructure to manage/move that asset. This is not about “ownership” for the sake of owning but about composing financial logic the way you write software.

  2. The "Self-Healing" Payment (Intelligence): Legacy payments are reactive—you investigate after a decline. Instead sovereign intelligence in payments will be proactive, scanning the mempool data and evaluating gas prices and liquidity depth across Layer-2s before a transaction is broadcast. It is a payment that optimises its own success rate without human intervention.

  3. The Agent: In this era, the entity making financial decisions can also be a digitally native and autonomous AI Agent. As these agents take over the more "mundane" tasks of corporate treasury, these digital-native entities require an intelligence layer to function as their "eyes and ears" on the ledger.

Why Intelligence is so critical for Fiat Payment Systems

In a mature TradFi payments market where the pipes (Visa, Mastercard, SWIFT etc) are static and the fees are entrenched, the main way to find new margin is to optimise the data about the money. This is the domain of companies like Pagos.ai (an angel investment of the Fabric team), which identifies why transactions fail. A simple example might be if a $500 B2C transaction fails, an intelligence layer tells the merchant precisely why: Was it a "Soft Decline" (the bank was busy, try again in two hours)? A "Hard Decline" (insufficient funds)? Or a "Processor Timeout"? 

In the TradFi payments world, we have long accepted a certain "failure tax" as the cost of doing business: In the B2C world, intelligence is a subset of conversion rate optimisation and addresses the “soft failure” of abandonment seen in unoptimised checkouts. Over fiat B2B rails (e.g. ACH or SWIFT), 15–20% of cross-border payments result in failures, costing the global economy approximately $118.5 billion annually. For a mid-sized firm, even a 2% error rate can translate into $500,000 in annual losses strictly from investigation fees, bank "repair" charges, and the wasted labour of finance teams manually untangling the mess.

Opportunities for the Application of Intelligence in On-Chain Payments

Blockchain has already gone a long way to solve the "3-day wait" as well as the "intermediary cost slice" but there remain several areas of application of raw data and intelligence to further deliver business performance, akin to where data and intelligence help on fiat rails:

Case Study: How Ramp Network Fixes Suboptimal Transactions

Being an insider to Ramp Network*, we saw first-hand how select Access providers have matured along the intelligence vector: Ramp Network recognized that "Access" was not just a legal/regulatory/integration problem, but also a UX friction problem. By shortening the user journey by 50% they addressed the "soft failure" of user abandonment. Suboptimal transactions often fail due to user friction during the card 3D Secure (3DS) process. Ramp offers one-tap purchases for returning users via Apple Pay and saved cards, which drastically reduces cart abandonment. KYC is also a UX problem and Ramp unlocks access across several platforms with one pass success.

In addition, by intelligently routing to a cheaper chain, they can preserve the transaction vs facing abandonment due to gas fees. With analytics they guide customers towards better exchange rates by moving transactions from just below rate thresholds to a level above it. 

In common with modern fiat gateways, Ramp Network leverages its infrastructure to route transactions through the most compatible processors based on the user's location and card type to improve authorisation rates. Ramp can select the specific acquiring bank or card processor most likely to approve a transaction in each region (e.g., using local acquiring). Taking this one step further by evaluating variables in real time including the user's location, the specific bank’s "crypto-friendliness" score, and historical conversion rates, if one "pipe" was blocked by a conservative bank, the money could still flow through another. Worst case, Ramp Network doesn't just show an error; it detects bank-specific anti-crypto policies and immediately prompts the user for an alternative card or an instant bank transfer, saving the revenue in real-time. 

Many bank card issuers automatically still ultimately decline "crypto" transactions. Ramp works to ensure transactions are processed with the most accurate and "acceptable" merchant codes to the banking network, reducing the estimated 50% industry-wide failure rate for fiat-to-crypto purchases (even after KYC). The on-chain on-ramp industry is starting to mature to properly reflect a holistic perspective of costs – not just fiat-crypto swap costs but also including conversion success and lost customer flow.

Case Study: Xweave: Dealing with the intricacies of local orchestration integrations‍

‍
Founded by an experienced long time payments operator (and ex- colleague of a Fabric team member at PayPal in Asia), it was natural founder/market fit for Xweave* to have an implicit focus on how to optimise the integrations that make on-chain orchestration work with fiat ramps in the more exotic currency markets of Asia: Xweave has rolled out a dashboard providing real-time visibility into total transaction count and volume, enabling better monitoring, performance analysis, and decision-making. They have focused on transaction success by optimising for call backs and “hung” status on transfers, pre-transaction validations to proactively catch issues before execution (e.g., account format, compliance checks) as well as in-flight  rate-handling, including dynamic adjustments to accommodate rate expiration constraints set by VASPs. Xweave’s focus on these details shows that outside the bulk of volume in USD stablecoin orchestration, the devil is indeed in the detail in order to deliver end-to-end reliability especially for local currency payments. 

‍

The Machine Economy Tipping Point

We get fair pushback that the above "Pagos” level of data analysis and intelligence only becomes valuable on-chain when a company is doing millions of transactions there. At that scale, a 0.1% optimisation in gas or slippage becomes meaningful.  Other fair pushback has been that there is not a deep enough "data lake" of failures to optimise against anyway.  So, what might get us to this scale? Here are some possibilities:

  1. When volumes increase:  Such as when micro-payments or high-frequency B2B (e.g., automated supply chain streaming) payments become the norm on-chain. Then the cost of gas/slippage per second makes intelligence mandatory.  Companies such as Superfluid*, have been pioneering streaming settlement on-chain across several pioneering use cases in payroll and rewards. 
  1. When on-ramp margins become too compressed: At this point, a payment provider must then find the next 1–2 bps of margin, again causing a focus on intelligence. In 2024, the "on-ramp tax" was 300–500 bps (3-5%). In 2026, competition has already compressed this to 50–100 bps.
  1. When network uptime hits "Five Nines" (99.999%): At this point the only failures left are strategic (wrong route, bad timing, poor liquidity). This is the exact moment intelligence becomes the most valuable tool in the stack.
  1. The age of AI agents: An AI machine to machine agent can execute thousands of micro-transactions for API access. For these "bots," intelligence is a functional requirement because agents require programmable guardrails. They need an intelligence layer that handles a variety of logic or exceptions such as gas price spikes, retries across transactions across Layer-2s, or compliance with ISO 20022 for corporate tax reporting.  When an agent proposes a trade, the recipient's intelligence layer could perform a variety of “pre-flight” checks such as the provenance of the funds, the credentials of the bot, the estimates of the execution costs and time. This type of active intelligence can prioritise or reject various transactions. 


In a world where we take the access and connectivity for granted, value will shift to:

Companies such as AIsa.one (an angel investment by Fabric team members) are pioneering aggregated API resource access packaged with guardrails, scalable payment infrastructure for them as well as performance analytics for insights.

Achieving Operational Resilience for On-Chain Payments

Regardless, any push back relating to a lack of appropriate scale of payment volume on-chain today, should not cause us to pause focusing on data and intelligence applied to payments on-chain because most immediately ahead of us we face a massive B2B opportunity. In B2B payments there is a real pain to solve for and it's large. Much larger than remittances. For example, globally, an estimated $11.6 billion in business working capital is trapped in "in-flight" status at any given moment. This prevents companies from reinvesting in inventory or R&D. Payment failures can freeze millions in working capital and damage critical supplier relationships. Zip has some scary statistics of the hidden costs in Tradfi B2B payments today.

In shifting B2B payments on-chain we face larger average transaction amounts which of course accentuates addressing performance issues such as slippage, but within B2B the primary friction points of adoption are actually not related to access or performance of the blockchain but instead concern aspects of the holistic transaction such as compliance, reconciliation, verifiable counterparties and treasury control:

‍

‍

Additional Business Opportunities from On-Chain Payments Intelligence

Data Intelligence for payments on-chain can also open new tangential fields which we are familiar with from Tradfi payments:

‍

Payment Incumbents are Not Standing Still

Visa, as a leading example of one of these forward looking incumbent payment players, has been very active in blockchain initiatives to date to ensure they integrate stablecoins and support tokenised assets to position it as the relevant payment layer for stablecoin payments – whether that’s 24/7/265 banking settlement with USDC/Solana, stablecoin-linked cards, stablecoin analytics or Visa Direct’s stablecoin prefunding for liquidity management or payouts to stablecoin wallets.

Card fees continue to be referenced as the low hanging target by the blockchain industry. With CEDP, this has started to change. At CEDP’s introduction in April 2026, interchange will be lower. US business credit (note not debit) card interchanges between merchants who are verified vs not, will be a significant difference between 1.75% and 2.65%. Of course, even if 1.75% may seem high vs blockchain, but we cannot forget that there remains a considerable value proposition offered by Visa beyond transfer of funds reflected in their fees. This ranges from refunds, chargebacks, operational reliability/fall-overs, legal certainty and governance (the importance was discussed above), liquidity via the 4-party model, security (although institutional MCP might be the alternative blockchain solution here) and trust (although with ZKP and other on-chain privacy technologies, trust-lessness could become acceptable for B2B participants on-chain). Importantly for me, two additional value props come with the card networks, both being crucial for B2B adoption on-chain: A) credit and b) payments data.

A) Credit cards offer a critical working capital life blood, especially for small businesses. Currently B2B buyers typically use cards to extend Days Payables Outstanding (DPO), effectively getting a 30-day interest-free loan. Instant stablecoin settlement is often a disadvantage to a buyer because it eliminates this "free float”. So, whilst blockchain still has an inherent fee advantage over even adjusted CEDP Visa interchanges, sellers should be encouraged to weaponise this delta of ~2.4% savings to incentivise buyers, e.g. offering a 1.5% discount for stablecoin payments to compensate for lost DPO.

B) On top of the fee reductions delivered by CEDP, it is much more a masterclass in using AI to defend a legacy intelligence throne: Under the old system, Visa merchants could often qualify for lower fee rates simply by meeting minimal data requirements, even if the included fields didn’t meaningfully describe a given transaction. CEDP removes that flexibility entirely by raising the bar on data and future proofing it.

For blockchain to truly solve B2B, it must move beyond the "cheap pipe" narrative and match the data and intelligence layer Visa has built for the age of agentic commerce.  Stablecoin payments practitioners must take up this mantle and force a choice between a legacy system that asks for data and a new system where the money and the data are mathematically one.

‍

2030 Vision: The Sovereign Intelligent Payment-Chain

By 2030, the historic Access phase will be viewed equivalently to the "Internet dial-up era” of on-chain payments. Looking forward payments on-chain will become self-remedial and self-optimising. We will see streaming settlements make the "Net-30" invoice extinct. As a ship crosses a GPS geofence, smart contracts will trigger value streams that agents instantly sweep into yield protocols, so value follows lockstep with physical goods. Every dollar moved will carry sovereign data and logic with it, which intelligence layers use to verify trust, adhere to policies (including pricing) and execute related workflow processes.  This is the Machine-to-Machine world where AI agents negotiate, settle, and report trades, with no human intervention, on performant self-optimising chains. 

This is the world we at Fabric Ventures invest into.

* Denotes an example company which is a Fabric Ventures investment.