This week’s reports point to a clear shift in banking and fintech: the industry is moving from legacy infrastructure and manual workflows toward real-time, AI-driven, and programmable systems. Across stablecoin payments, agentic banking, and new operating models, the message is consistent — banks that modernize their technology stack, embed automation into execution, and build around trust and compliance will be best positioned to grow. At the same time, Europe’s fintech ecosystem, virtual cards, and rising VC activity show that capital is increasingly flowing toward infrastructure, efficiency, and on-chain finance.

Video of the Week

Deep Dive of the Week

The Emergence of the Bank Operating System

For decades, the financial industry has operated under the assumption that the core banking system is the sun around which all other technologies must orbit. This core-centric model was functional in an era of physical branches and daily batch cycles, but it has become a terminal constraint in the modern digital economy. The core is a monolithic system of record designed for stability over velocity. It was built to store data, not to orchestrate real-time commerce. In my analysis, the gravitational pull of these legacy cores is currently preventing banks from participating in the most lucrative segments of the fintech revolution.

Every modernization strategy over the last twenty years has been a variation of architectural accumulation. Banks add middleware to translate legacy code. They add point solutions to handle specific mobile features. They add third-party Banking as a Service platforms to enable fintech partnerships. The result is a surface-level digital experience that masks an underlying foundation of legacy debt. Smaller institutions that remain release cycle-bound, unable to launch a single feature without their vendor’s permission. Mid-sized banks are struggling with fragmented modernization and integration drag. Large banks are attempting in-house builds that are too expensive and slow to scale. These are different strategies, but they all suffer from the same friction.

What I see now is the formation of a new category: the Bank Operating System. This is a real-time, composable growth layer that repositions strategic control inside the bank. In this model, the core becomes the system of record while the operating system becomes the system of growth. This distinction is foundational. A real-time operating system layer helps banks launch revenue-generating products quickly, activates new payment rails without integration friction, and operates independently of core downtime.

This week’s reports

1️⃣The Agentic Banking Blueprint

2️⃣Technology Trends Redefining Banking

3️⃣State of European FinTech

4️⃣Stablecoin Payments Infrastructure

5️⃣Fintech VC Trends Q1 2026

6️⃣State of Fintech Q1’26 Report

7️⃣Virtual Cards as a Strategic Advantage

The Agentic Banking Blueprint

Over the past two decades, most transformation efforts have focused on digitization. Banks invested heavily in systems to capture data, streamline processes, and improve digital customer experiences. More recently, artificial intelligence (AI) has enhanced these efforts by helping employees analyze information, generate insights, and make better decisions.

However, the way work gets done has not changed as much as expected. Most execution still depends on manual coordination across teams, systems, and channels. Even when insights are available, follow-up is often inconsistent. Processes slow down where decisions need to be turned into action.

AI agents address this gap. Unlike traditional AI tools that only provide recommendations, AI agents are designed to take action. These agents can complete defined tasks, follow rules, and move processes forward across systems.

Now, a more advanced model is emerging: autonomous AI agents. These agents represent a new era of intelligent automation; they can learn from data, make decisions within defined objectives, and adapt to real-world conditions. Rather than just executing tasks, they manage outcomes within established guardrails and continuously improve performance over time.

Read more

For decades, the financial industry has operated under the assumption that the core banking system is the sun around which all other technologies must orbit. This core-centric model was functional in an era of physical branches and daily batch cycles, but it has become a terminal constraint in the modern digital economy. The core is a monolithic system of record designed for stability over velocity. It was built to store data, not to orchestrate real-time commerce. In my analysis, the gravitational pull of these legacy cores is currently preventing banks from participating in the most lucrative segments of the fintech revolution.

Every modernization strategy over the last twenty years has been a variation of architectural accumulation. Banks add middleware to translate legacy code. They add point solutions to handle specific mobile features. They add third-party Banking as a Service platforms to enable fintech partnerships. The result is a surface-level digital experience that masks an underlying foundation of legacy debt. Smaller institutions that remain release cycle-bound, unable to launch a single feature without their vendor’s permission. Mid-sized banks are struggling with fragmented modernization and integration drag. Large banks are attempting in-house builds that are too expensive and slow to scale. These are different strategies, but they all suffer from the same friction.

What I see now is the formation of a new category: the Bank Operating System. This is a real-time, composable growth layer that repositions strategic control inside the bank. In this model, the core becomes the system of record while the operating system becomes the system of growth. This distinction is foundational. A real-time operating system layer helps banks launch revenue-generating products quickly, activates new payment rails without integration friction, and operates independently of core downtime.

The architectural debt crisis

90% of US banking core software is considered legacy. These systems were designed for a different era where batch processing and rigid architectures were the norm. In a batch environment, transactions are stored and processed in groups, usually overnight. This is incompatible with a world that demands instant payment confirmation and 24/7 availability.

The true cost of staying on these legacy systems is staggering. I have seen data indicating that universal banks allocate 70% of their IT budget to system maintenance, leaving only 30% for innovation. As the pool of COBOL and mainframe experts shrinks due to retirement, the cost of maintaining these systems escalates. Finding specialists who understand these decades-old systems is becoming difficult and expensive every year. When a bank’s primary technology is a black box that few people understand, the risk of a catastrophic failure increases every year.

Middleware was supposed to solve this problem by layering modern experiences on top of legacy cores. I believe this has only provided a temporary solution. Middleware is still mostly a connective tissue to the core and continues to depend on fragmented, older systems for functionality, which complicates integration and creates performance issues. Middleware does not solve the fundamental problem of siloed data and batch constraints. It merely hides them. Banks using this approach still suffer from fragmented experiences across mobile, web, and branch channels. Customers expect real-time visibility, and they are increasingly frustrated when a mobile app shows one balance while a branch teller sees another.

Defining the Bank Operating System category

I define a modern bank operating system as a cloud native environment that surrounds legacy cores to unify all banking capabilities. It is a modular system that enables banks to innovate at the speed of fintech by leveraging modern, composable technology architecture alongside Artificial Intelligence. Unlike a traditional core, which is focused on the storage of accounting records, a bank operating system is focused on the execution of business logic and money movement. It serves as an innovation core that allows the bank to move fiat currency, tokenized deposits, and other digital assets across various networks from a single environment.

Six unified pillars form the foundation of a modern BankOS.

• The first is Unified Banking, which uses a single modern ledger and a unified payments engine to handle both fiat and tokenized payments.

• The second is Unified Experience, abstracting the complexity of underlying rails.

• The third is Unified Payments, which consolidates ACH, Fedwire, RTP, FedNow, SWIFT, and Blockchain on one platform.

• The fourth is Unified Operations, delivering a single view across channels, assets, and rails, with AI-driven reconciliation and compliance.

• The fifth is Unified Integration, which is a layer that allows plug-and-play connectivity into any ecosystem.

• Finally, Unified Intelligence provides predictive analytics and deep insights into customer behavior.

This architecture repositions the bank’s strategic control. Historically, innovation was gated by the release cycles of core vendors. Under a bank operating system model, the operating system becomes the control plane. This allows the bank to become an ecosystem participant rather than just a utility provider. It enables embedded banking and sponsor models through configuration instead of re-architecture. Transformation shifts from a high-risk infrastructure project to a revenue acceleration strategy.

The technical mechanics of the sidecar core

I am a proponent of the sidecar core strategy because it offers a low-risk path to modernization. A sidecar core is a separate, modern banking system that runs parallel to legacy infrastructure. It does not require a full replacement of the traditional core. Instead, it gradually takes over customer segments, products, or services, becoming the mainstream core and giving banks a clear exit from legacy systems. I have seen projections that 40% of global banks will pursue sidecar strategies by 2026, rising to 80% by 2028.

The technical architecture of a sidecar core, such as Finzly BankOS, relies on a smart next-gen parallel ledger. It connects to the legacy core and third-party systems using adapters and an extensive suite of APIs. I find that this approach allows for non-disruptive integration. The sidecar core handles the real-time demands of modern payments and deposits, while the legacy core continues to manage bulk operations for the existing customer base. This creates a migration runway where entire portfolios can shift naturally over months instead of years.

A sidecar core must be ISO 20022 native to be effective. ISO 20022 is the global messaging standard that enables data-rich transactions. Legacy cores typically rely on flat files and rigid schemas that truncate data. An ISO 20022 native bank operating system can ingest 24/7 transactions and overcome the batch constraints of the core. This is essential for participating in real-time networks like FedNow and RTP, which require millisecond journaling that old mainframes cannot achieve.

Real-time payments and the A2A imperative

I believe the proliferation of account-to-account payments and digital wallets is the most significant pressure vector forcing banks to adopt operating systems. Digital wallets are multiplying across geographies, and digital-first customers expect instant funds availability. Banks that cannot flexibly integrate A2A rails and wallet networks risk being disintermediated from payment flows they once controlled. Fragmented vendor stacks cannot keep pace with the velocity of this market.

A unified operating system allows a bank to activate new rails as configuration events. For example, a bank operating system can unify ACH, Fedwire, RTP, FedNow, SWIFT and even on-chain rails into a single payments processing platform. This provides total control and visibility across all processors from one API, one customer experience or one operational console. It eliminates the need for separate integrations for each payment type. This is critical for US instant payment rails like FedNow and RTP, which settle in real time 24/7. Legacy batch cores simply struggle to maintain balance accuracy in an always-on environment.

There is also a structural shift in how businesses manage liquidity. Digital adoption has increased the demand for real-time notifications and automated reconciliation. An operating system purpose-built for real-time operation enables continuous engagement and instant decision-making. This allows banks to compete with fintechs by offering instant payment confirmation and immediate settlement. Here, speed to value is a competitive metric that is now a balance sheet variable. The institutions that reduce time to market reduce their strategic risk.

Virtual accounts and programmable liquidity

There is a clear demand for smarter, more flexible account structures. Modern financial ecosystems require virtual accounts layered over master accounts, multi-entity platform structures, and segmented balances for marketplaces. These are not edge cases; they are standard requirements for corporate clients today. The composable ledger on the bank operating system enables these programmable account hierarchies without forcing a core re-architecture. This unlocks new deposit models and monetizable partner offerings.

Technically, virtual accounts are non-real accounts used to redirect transactions to an underlying real account. They are off-balance sheet items from the bank’s perspective. I have seen how multiple virtual accounts can be mapped to a single real account to help a corporate client rationalize their cash positions. Composable virtual accounts support single-level or multi-level hierarchical structures. Transactions performed on any virtual account are automatically routed through the hierarchy in real time, while shadow entries are posted to the respective virtual accounts.

This structure is liberating for treasurers. It allows for the notional segregation of funds through shadow transaction postings, aiding in automated reconciliation. Corporations use these structures to mimic in-house banks or centralized payment factories. For banks, providing these capabilities through an operating system means they can onboard fintech partners at scale and manage complex ownership structures like escrow or IOLTA accounts without adding operational complexity to the legacy core.

The shift to agentic AI in banking

I want to distinguish between traditional automation and the new wave of Agentic AI. For years, banks have used Robotic Process Automation to handle repetitive, rule-based tasks. RPA is like an assembly line worker following exact instructions. It is deterministic and brittle. If a UI element moves or a label changes, the bot breaks. Agentic AI is fundamentally different. It is goal-driven and adaptive. It uses large language models and reasoning to pursue outcomes through strategic decision-making.

Unlike generative AI, which just creates content, Agentic AI takes action. I see agents that investigate transaction disputes, resolve exceptions autonomously, and coordinate multi-step workflows across disconnected systems - but always with human oversight to ensure safe decision-making in financial services. A modern operating system with agents can automate payment operations and compliance by embedding intelligence directly into the payment flow. For example, an AI agent can analyze a wire flag, pull context from account history and geographic data, and recommend a resolution in seconds.

I believe this shift from rule following to reasoning enables true end-to-end automation. AI agents do not break when vendor formats change or ERPs get upgraded. They adapt. Data suggests that AI agents can reduce operational costs by 30 to 40% by 2030. They free up human talent to focus on high-value advisory roles. In an operating system, the back office becomes a digital engine that supports efficiency and superior customer outcomes.

The tokenized economy

From my latest conversation, I am observing a significant strategic shift toward tokenized assets. The market for stablecoin issuance is projected to reach 1.9 trillion dollars by 2030. Banks face a choice: they can either participate in this ecosystem or watch as liquidity moves to fintech competitors. Tokenized deposits are digital representations of regulated bank liabilities that stay within the traditional banking system. Stablecoins, conversely, are typically issued by private entities and backed by reserves. Both are digital money tokens that operate on 24/7 rails with programmability.

A bank operating system provides the infrastructure to manage these assets without betting on a single blockchain. I see modules like Finzly’s Token Galaxy as a way to unify traditional and tokenized rails. This allows banks to move fiat, tokenized deposits, and stablecoins across networks through a multi-asset, multi-rail engine on the operating system, with operational oversight from a single dashboard. This interoperable infrastructure is essential for strategic freedom. It allows banks to connect to multiple blockchains and consortia simultaneously. It also enables advanced use cases like automatic payroll deposits into tokenized accounts.

The technical foundation of this tokenization layer must be robust. Multi-Party Computation secured wallets ensure enterprise-grade infrastructure with no single point of failure. Tokenized movements must meet the same rigorous KYC and AML standards as fiat systems. Composable workflows integrate compliance directly into event-driven, multi-asset money movement, providing real-time monitoring and a complete audit trail. This is a way for banks to retain deposits and prevent attrition to unregulated digital asset platforms.

Speed to value as a balance sheet variable

The primary metric for modern banking should be Speed to Value. Historically, banks have focused on Total Cost of Ownership, but this does not account for the opportunity cost of slow innovation. Speed to Value measures the velocity at which an institution can move from a business idea to revenue generation. In a dynamic market, the ability to launch a product in weeks rather than years is a significant competitive advantage. Cloud migration and low code adoption are key enablers of this velocity.

Modernizing infrastructure is no longer just about reducing IT costs; it is about revenue acceleration. The bank operating system model transforms the ROI calculations. Instead of multi-year infrastructure projects with uncertain outcomes, banks can deploy modular capabilities and measure immediate financial impact. For example, some institutions use sidecar cores to launch new digital channels and increase digital adoption by 80%. This is the ability to create capacity in previously constrained parts of the organization.

Also, Speed to Value has a direct impact on bank valuations. By improving EBITDA and increasing the growth multiple, digital transformation creates measurable enterprise value. Analysts are increasingly focusing on KPIs like customer acquisition cost, retention rate, and revenue growth rate. A bank operating system that enables rapid product iteration and superior customer experience directly influences these metrics. I believe the standard procurement and implementation cycle of 6 to 9 months is being replaced by a model that prioritizes iterative releases and continuous value realization.

The future of the bank operating system

I believe we are entering a phase where the core becomes invisible. In the future, the primary interface for both the bank and the customer will be the operating system. The modern bank operating system is evolving into a fully autonomous environment where AI agents handle the majority of operational tasks. This will lead to a touchless support model, still with human oversight, and fully automated customer data management. The competitive advantage will go to those who have the best data foundation and the most integrated systems.

The integration reality check is that while most firms have some digital adoption, few have full system integration. The firms with fully integrated systems report 10x better growth performance and ROI measured in weeks. The biggest barrier to AI effectiveness is fragmented data. The most successful firms focus on integration over addition. They connect their CRM with core banking, payments, and compliance tools to create a single source of truth. This is a foundational requirement for any bank that wants to survive the next decade.

I believe that the bank operating system is the structural shift that will define the next era of banking. It reframes modernization from a defensive move to an offensive revenue strategy. By decoupling the innovation cycle from the legacy core, banks can participate in the real-time, tokenized economy. Operating systems like Finzly BankOS are becoming the control plane for the future of finance, enabling banks to provide the personalized, instantaneous services that modern customers demand. The institutions that embrace this model will thrive; those that remain core-centric will obsolesce.

Sources:

1. BankOS Platform - Finzly, accessed April 5, 2026, https://finzly.com/platform/

2. How Banks Are Layering Modern Experiences on Top of Legacy Cores - Veriday, accessed April 5, 2026, https://veriday.com/modernizing-banking-experience-without-core-replacement/

3. Learn how progressive modernization enables banks to transform core systems without big bang risk. - Softjourn, accessed April 5, 2026, https://softjourn.com/insights/core-banking-modernization-in-5-steps

4. Balancing Cost and Risk with Growth and Revenue, accessed April 5, 2026, https://arizent.brightspotcdn.com/18/5b/080f452b482fa9a83fb774c2bde8/bank-modernization-ebook-v6.pdf

5. Legacy Banking Systems Explained: Why Modernization Matters - Archon Data Store, accessed April 5, 2026, https://www.archondatastore.com/blog/legacy-banking-system-modernization/

6. what is a modular bank? | Skaleet, accessed April 5, 2026, https://skaleet.com/en/blog/modular-bank-what-is-a-modular-bank

7. Banking Operating System: Core to Platform Guide - Backbase, accessed April 5, 2026, https://www.backbase.com/blog/banking-operating-system

8. Finzly BankOS.pdf

9. Tokenized Deposits & Stablecoins Platform for Banks | Finzly, accessed April 5, 2026, https://finzly.com/solutions/tokenized-deposits/

10. Finzly Unveils Token Galaxy: Unifying Money Movement for Traditional and Tokenized Banking - PR Newswire, accessed April 5, 2026, https://www.prnewswire.com/news-releases/finzly-unveils-token-galaxy-unifying-money-movement-for-traditional-and-tokenized-banking-302710960.html

11. Inside Finzly’s Vision: Rethinking Banking Infrastructure with AI and ..., accessed April 5, 2026, https://finzly.com/resources/blogs/inside-finzlys-vision-rethinking-banking-infrastructure-with-ai-and-tokenized-money/

12. Building on Bedrock: A Practical Guide to Core & Ledger Choices in Embedded Banking, accessed April 5, 2026, https://ingomoney.com/blogs/building-on-bedrock-a-practical-guide-to-core-ledger-choices-in-embedded-banking/

13. Sustainable Scaling: Why Modularity Makes Sense for LatAm Banks in 2026, accessed April 5, 2026, https://www.galileo-ft.com/blog/sustainable-scaling-why-modularity-makes-sense-for-latam-banks/

14. Stablecoins: Exploring the Future of Cross-Border Payments - Finzly, accessed April 5, 2026, https://finzly.com/resources/blogs/stablecoins-exploring-the-future-of-cross-border-payments/

15. Finzly India R&D is hiring Fullstack Developer job in Pune, Chennai | Cutshort, accessed April 5, 2026, https://cutshort.io/job/Fullstack-Developer-Pune-Chennai-Finzly-India-RD-LNmjTUP3

16. Low-code platforms deliver high rewards - KPMG International, accessed April 5, 2026, https://kpmg.com/pt/en/insights/2023/06/low-code-platforms-deliver-high-rewards.html

17. Accelerate Your Cloud Journey with Altimetrik, accessed April 5, 2026, https://www.altimetrik.com/blog/ai-first-cloud-journey-acceleration/

18. Virtual Accounts - Enabling lean treasuries - Whitepaper | Oracle, accessed April 5, 2026, https://www.oracle.com/a/ocom/docs/industries/financial-services/fs-vam-enabling-lean-treasuries-wp.pdf

19. Empowering Banks With Virtual Ledgers - Datos Insights, accessed April 5, 2026, https://datos-insights.com/reports/empowering-banks-with-virtual-ledgers/

20. Agentic AI vs RPA: Practical Differences, Real Use Cases, and How to Combine Them, accessed April 5, 2026, https://www.mywave.ai/blog/agentic-ai-vs-rpa

21. RPA vs. AI Agents for Accounting: Which is Better for Modern Finance Teams?, accessed April 5, 2026, https://blog.auditoria.ai/rpa-vs-ai-agents-for-accounting-finance

22. AI agents versus RPA: A guide for accountants - Thomson Reuters, accessed April 5, 2026, https://tax.thomsonreuters.com/blog/ai-agents-versus-rpa-a-guide-for-accountants-tri/

23. 7 Use Cases For Agentic AI in Banking - Druid AI, accessed April 5, 2026, https://www.druidai.com/blog/7-use-cases-for-agentic-ai-in-banking

24. Finzly Announces Agentic AI-Powered Payments and Operations, accessed April 5, 2026, https://finzly.com/resources/press-releases/finzly-announces-agentic-ai-powered-payments-and-operations/

25. How Retail Banks Can Put Agentic AI to Work | BCG, accessed April 5, 2026, https://www.bcg.com/publications/2026/how-retail-banks-can-put-agentic-ai-to-work

26. How is Agentic AI different from RPA? - AutomationEdge, accessed April 5, 2026, https://automationedge.com/blogs/how-is-agentic-ai-different-from-rpa/

27. Tokenized Deposits vs. Stablecoins: A Practical Guide for Financial Institutions | Stablecore Blog, accessed April 5, 2026, https://stablecore.com/blog/tokenized-deposits-vs-stablecoins-a-practical-guide-for-financial-institutions

28. Stablecoins and tokenized deposits are poised to reshape global finance | Deloitte Canada, accessed April 5, 2026, https://www.deloitte.com/ca/en/Industries/financial-services/perspectives/tokenized-money-stablecoins.html

29. contest between central bank digital currencies, stablecoins, and tokenized deposits: Which will likely win, and why? | Capital Markets Law Journal | Oxford Academic, accessed April 5, 2026, https://academic.oup.com/cmlj/article/21/2/kmag012/8571935

30. Finzly Unveils Token Galaxy: Unifying Money Movement for Traditional and Tokenized Banking, accessed April 5, 2026, https://finzly.com/resources/press-releases/finzly-unveils-token-galaxy-unifying-money-movement-for-traditional-and-tokenized-banking/

31. Tech implementation sets your speed to value | Grant Thornton, accessed April 5, 2026, https://www.grantthornton.com/insights/articles/advisory/2024/tech-implementation-sets-your-speed-to-value

32. KPMG Elevate - Value Creation, accessed April 5, 2026, https://kpmg.com/cn/en/services/advisory/value-creation.html

33. Best financial services software for banks of March 2026 - FitGap, accessed April 5, 2026, https://us.fitgap.com/search/financial-services-software/banks

34. Commercial banking platforms for SMBs & mid-market 2024 - Backbase, accessed April 5, 2026, https://www.backbase.com/blog/top-commercial-banking-platform-providers

35. Modernizing the Banking Ledger: Powering Smarter Core Banking - Vacuumlabs, accessed April 5, 2026, https://vacuumlabs.com/articles/banking-ledger/

36. Top AI-native banking platform providers in 2026: a buyer’s guide - Backbase, accessed April 5, 2026, https://www.backbase.com/blog/top-ai-native-banking-platform-providers

37. Core Banking Platforms Compared: Features, Pricing, and Integration Ease | Gemba, accessed April 5, 2026, https://ge.mba/research/core-banking-platforms-compared-features-pricing-and-integration-ease

38. Agentic AI Integration: Business Process Automation Guide - Informatica, accessed April 5, 2026, https://www.informatica.com/resources/articles/agentic-ai-integration.html

39. HubSpot Asia Pacific | Singapore Finance Firms: Why Only 30% See ROI in Weeks From Digital Transformation, accessed April 5, 2026, https://www.hubspot.com/apac/newsroom/blog-singapore-financial-services-roi

Disclaimer:

Fintech Wrap Up aggregates publicly available information for informational purposes only. Portions of the content may be reproduced verbatim from the original source, and full credit is provided with a “Source: [Name]” attribution. All copyrights and trademarks remain the property of their respective owners. Fintech Wrap Up does not guarantee the accuracy, completeness, or reliability of the aggregated content; these are the responsibility of the original source providers. Links to the original sources may not always be included. For questions or concerns, please contact us at sam.boboev@fintechwrapup.com.

One idea stayed with me throughout the conversation: many fintechs optimize the surface layer, but few are willing to rebuild the foundation. Shannon explained why Airwallex…

Read more

The reports highlight a major shift in financial infrastructure driven by agentic AI, stablecoins, and automation. Payments are moving from human led actions to agent executed transactions across three flows: human to agent, agent to business, and agent to agent. This shift introduces new challenges around authorization, trust, and liability.

Stablecoins are also evolving into core financial rails. Most activity is now focused on B2B payments, liquidity management, and cross border settlement. The market is consolidating around compliant and institution friendly players, positioning stablecoins as foundational infrastructure rather than speculative assets.

In banking, agentic AI is transforming KYC from a manual and costly process into a more automated system where AI handles routine tasks and humans focus on exceptions. This can significantly improve efficiency and reduce costs.

Overall, financial systems are becoming more automated and interoperable, with AI agents and programmable money reshaping how transactions are initiated, processed, and governed.

Video of the Week

Deep Dive of the Week

The Ultimate Directory of Agentic Commerce and Payment Protocols / Tools

What this directory covers

“Agentic commerce” is a shift from human-operated checkout (clicking “Buy”) to software agents that discover, negotiate, and transact on a user’s behalf often across multiple systems, surfaces, and payment methods. Multiple initiatives are trying to prevent this ecosystem from fragmenting into one-off integrations by defining shared languages, primitives, and trust/authorization artifacts.

This directory treats “protocol” broadly as any standardized interface that enables interoperable agent-led commerce, including:

• Commerce/checkout standards (how agents and merchants coordinate carts, checkout, order lifecycle).

• Payments authorization and audit standards (how intent, consent, and accountability become cryptographically provable).

• Trust and traffic authenticity mechanisms (how merchants distinguish trusted commerce agents from malicious automation).

• Agent tooling surfaces (MCP servers, SDKs, toolkits) that make these protocols practical to implement.

This week’s reports

1️⃣The Shift to Agentic Payments

2️⃣The Definitive Stablecoin Landscape Series

3️⃣Case study - Klarna

4️⃣The Know-Your-Customer Agentic AI Revolution

5️⃣The $340 Million Case for Tokenized Collateral

6️⃣How Revolut Built PRAGMA, an AI Foundation Model

7️⃣The Global Payments Report 2026

The Shift to Agentic Payments

Commerce is evolving through a sequence that is familiar across many technology cycles.

Initially, humans transacted directly with each other. Eventually, software embedded payments inside applications and digital products. Now, agents are becoming the actors that decide when and how transactions occur.

That shift creates three distinct emerging payment flows:

-> Human-to-agent. A person delegates authority to an agent within defined constraints. For example, an AI assistant books travel within a budget, renews subscriptions, or completes routine purchases on a user’s behalf. In these cases the core challenge is safe delegation, where systems must determine who authorized the action, what spending limits apply, and where liability sits when software acts on behalf of a user.

-> Agent-to-business. Agents interact directly with merchants, vendors, or APIs without a human involved in the transaction itself. This could be an AI agent purchasing compute from a cloud provider, paying for an API request, or ordering supplies within an enterprise procurement system. In these cases payments become part of the operational workflow rather than a checkout step.

Read more

What this directory covers

“Agentic commerce” is a shift from human-operated checkout (clicking “Buy”) to software agents that discover, negotiate, and transact on a user’s behalf often across multiple systems, surfaces, and payment methods. Multiple initiatives are trying to prevent this ecosystem from fragmenting into one-off integrations by defining shared languages, primitives, and trust/authorization artifacts.

This directory treats “protocol” broadly as any standardized interface that enables interoperable agent-led commerce, including:

• Commerce/checkout standards (how agents and merchants coordinate carts, checkout, order lifecycle).

• Payments authorization and audit standards (how intent, consent, and accountability become cryptographically provable).

• Trust and traffic authenticity mechanisms (how merchants distinguish trusted commerce agents from malicious automation).

• Agent tooling surfaces (MCP servers, SDKs, toolkits) that make these protocols practical to implement.

  Fintech Wrap Up is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Protocol directory

Universal Commerce Protocol (UCP)

UCP positions itself as an open, standardized “common language” for platforms/agents and businesses to interoperate across the commerce lifecycle, explicitly spanning discovery, buying, and post-purchase experiences.

On Google’s implementation track, UCP is framed as a way to turn AI interactions into direct buying, starting with direct purchase experiences across Google AI surfaces (the guide references AI Mode in Search and Gemini). It highlights two integration paths: native checkout (deeper agentic integration) and an embedded checkout option for approved merchants with complex flows.

In the open-source spec, UCP emphasizes:

• Composable capabilities and extensions (e.g., “Checkout,” “Order,” “Identity Linking”), rather than monolithic integrations.

• Transport flexibility, describing support for REST, MCP, and A2A (and explicitly referencing industry standards like JSON-RPC transports).

• Identity linking via OAuth 2.0 and compatibility/security alignment with AP2-style mandates and verifiable credentials.

UCP belongs in the “commerce interoperability” category: it aims to standardize how agents/platforms talk to merchants, while remaining compatible with a broader payments trust layer (AP2) and agent tooling patterns (MCP, A2A).

Read more

What stood out most from our conversation is that the future of payments may not be defined by a single new rail or feature, but by how trust, control, and simplicity c…

Read more

This week’s reports all point to the same shift: finance is moving toward a more programmable, real-time, and autonomous operating model. Tokenized deposits are presented as a banking infrastructure upgrade rather than a crypto replacement, while stablecoins and CBDCs are emerging as parallel rails in a more controlled monetary stack. At the same time, payments are being shown as fragmented across multiple systems, which makes observability and reconciliation a core challenge. The other major theme is AI agents: they are evolving from tools that recommend actions into actors that can search, negotiate, procure, and pay within defined limits. Taken together, the reports suggest that commerce and finance are converging into a new system where settlement, compliance, and execution happen closer together, and where banks, fintechs, and AI systems all play roles in the same evolving infrastructure.

Video of the Week

Deep Dive of the Week

An Analytical Breakdown of Revolut’s 2025 Performance

The fiscal year 2025 represents a landmark epoch for Revolut, marking a decade of operational existence and a definitive transition from a high-growth fintech disruptor to a globally systemic financial institution. The dual analysis of the firm’s 2025 Annual Report and the independent assessment by Andreessen Horowitz (a16z) reveals a financial profile characterized by “outlier” performance metrics that deviate significantly from both traditional banking incumbents and modern neobank peers. With a reported revenue of £4.5 billion, a 46% year-on-year increase, and a profit before tax of £1.7 billion, representing a 57% surge, Revolut has demonstrated a capacity for compounding efficiency that validates its technology-first operating model. This transition is underpinned by a “Growth Algorithm” that prioritizes product velocity, regulatory statecraft, and extreme operational leverage, resulting in a return on equity that challenges the fundamental assumptions of the consumer finance sector.

This week’s reports

1️⃣Tokenized Deposits Essential to the Future of Digital Finance

2️⃣Stablecoins and the Future of Payments: Evidence from Financial Markets

3️⃣From Click to Command: The Shift to Agentic Payments

4️⃣Building Gulf stablecoins and CBDCs infrastructures Strategy, Scale and Supervision (2026-2030)

5️⃣Let’s dive into 9 fintech predictions in 2026

6️⃣The Agentic AI Advantage: Finance agents that move the numbers

7️⃣Everything to Know About the New Stack for Autonomous Commerce

Tokenized Deposits Essential to the Future of Digital Finance

Most people still think tokenized deposits are a crypto story. But I would say they are a banking infrastructure upgrade.

Tokenized deposits are simply commercial bank money represented on a blockchain. The key difference is not what they are, but how they move. They enable 24/7, near real-time settlement with programmable logic, while remaining fully within the regulated banking system.

____

This is why banks are leading, not being disrupted.

Unlike stablecoins, tokenized deposits are direct liabilities of banks, backed by existing deposits, covered by deposit insurance, and governed by the same regulatory frameworks. They preserve the two-tier monetary system instead of bypassing it.

____

The shift is already operational, not theoretical.

J.P. Morgan, Citi, and BNY are running live systems for institutional clients. Corporates are using them for treasury automation, FX, and liquidity management. Transactions that used to take hours or days are executed instantly and programmatically across time zones.

____

The real change is structural.

-> Settlement is collapsing into the same layer as execution.

Read more

The fiscal year 2025 represents a landmark epoch for Revolut, marking a decade of operational existence and a definitive transition from a high-growth fintech disruptor to a globally systemic financial institution. The dual analysis of the firm’s 2025 Annual Report and the independent assessment by Andreessen Horowitz (a16z) reveals a financial profile characterized by “outlier” performance metrics that deviate significantly from both traditional banking incumbents and modern neobank peers. With a reported revenue of £4.5 billion, a 46% year-on-year increase, and a profit before tax of £1.7 billion, representing a 57% surge, Revolut has demonstrated a capacity for compounding efficiency that validates its technology-first operating model. This transition is underpinned by a “Growth Algorithm” that prioritizes product velocity, regulatory statecraft, and extreme operational leverage, resulting in a return on equity that challenges the fundamental assumptions of the consumer finance sector.

The financial architecture of scaled efficiency

Revolut’s 2025 financial results provide the empirical bedrock for its current $75 billion valuation, establishing it as the most valuable private technology company in Europe. The organization has achieved a rare equilibrium between aggressive top-line expansion and bottom-line profitability, entering its fifth consecutive year of net profit in 2025. According to a16z analysis, a critical component of this success is the “Rule of 75,” a metric formulated by combining revenue growth and net profit margin. For the 2025 period, Revolut achieved a revenue growth rate of 46% (57% when translated to USD) and a net profit margin of 29%, placing it in a unique tier of efficiency that few financial institutions, modern or established, have historically reached.

The composition of Revolut’s revenue streams highlights a diversification strategy that mitigates the interest rate sensitivity typically found in traditional banking models. While many established banks derive upwards of 70% of their income from interest, Revolut has inverted this structure, with fee-based revenues accounting for 76% of total turnover in 2025, an increase of 4.2 percentage points over the prior year. This diversification is realized across six primary segments, with no single category responsible for more than 22% of total revenue, ensuring that the firm is not reliant on a “one-trick pony” revenue model.

Revenue mix and segment performance

The 2025 revenue breakdown reveals that card payments remain the largest single revenue stream at 22.2%, closely followed by interest income at 21.6%. The growth in subscription revenue, which reached 15.7% of the total, is particularly significant as it represents a recurring, high-margin income source that reflects deepening customer engagement. Wealth services, including equities and cryptocurrency trading, contributed 14.7%, while foreign exchange services accounted for 13.4%. Internally, the company now tracks 11 distinct product lines that exceeded £100 million in revenue during the 2025 fiscal year, illustrating the breadth of its monetization engine.

The 2025 report also notes a significant shift in geographic revenue distribution. While the United Kingdom remains a core market, Revolut has achieved Euro-wide penetration, with no single country accounting for more than 25% of fee revenue. This regional resilience is bolstered by explosive growth in markets such as France, Italy, and Spain, where Revolut is increasingly perceived not as a secondary travel tool but as a primary bank account of choice. In Europe, approximately one in five working-age adults now uses Revolut, and in several key markets, nearly one in three newly opened bank accounts is a Revolut account.

The compounding algorithm: a16z perspectives on growth

Analysis from a16z posits that Revolut is one of the fastest-growing financial institutions in history, maintaining a remarkable 76% four-year Compound Annual Growth Rate (CAGR) in USD terms since crossing the $1 billion revenue threshold in 2022. This performance is particularly notable given that Revolut operates primarily in mature European markets with entrenched incumbents, unlike peers such as NuBank that have scaled in developing regions. The a16z study emphasizes that in 2022, Revolut had less revenue than many of its well-known fintech peers, including Robinhood, SoFi, and Chime; however, by 2025, it had surpassed them, generating up to three times the revenue of some competitors.

Read more

In this episode, I sit down with Michael Tannenbaum, the CEO of Figure, to break down what’s real and what’s misunderstood. We go beyond the headlines and get into how tokenization actually works in practice, starting with credit markets and expanding into equities.

We unpack why most tokenization efforts fail, what Figure is doing differently, and where the real bottlenecks sit today: not in technology, but in market structure, incentives, and fragmentation.

We also cover:

• Why liquidity does not magically appear when you “put assets on blockchain”

• The hidden inefficiencies in capital markets and who benefits from them

• Why debt markets may be a bigger opportunity than equities

• Tokenized vs blockchain-native assets and why the distinction matters

• Who wins and who gets squeezed as tokenization scales

• The real role of regulation and why it is not the biggest blocker

• What capital markets could look like in 5–10 years with AI and blockchain combined

This is a grounded conversation about how financial infrastructure actually evolves, not how we wish it would.

If you care about payments, capital markets, or where finance is heading next, this one is worth your time.

Read more

Video of the Week

Deep Dive of the Week

The Machine Payments Protocol

The financial infrastructure of the internet is being re-architected. On March 18, 2026, the launch of the Machine Payments Protocol (MPP) and the Tempo mainnet signaled the possible end of the human-centric checkout era. Actually, this could be a fundamental shift in how value moves across the web. For thirty years, the internet lacked a native value layer, forcing the industry to build increasingly complex user interfaces around legacy credit card rails. These interfaces, designed for human psychology, have become the primary bottleneck for the emerging agenticeconomy.y

I believe the fundamental friction of the modern web is the “human-in-the-loop” requirement for commerce. AI agents do not use browsers. They do not click buttons, they do not solve CAPTCHAs, and they do not possess the cognitive friction that leads to cart abandonment. When an AI agent requires a resource, payment is a programmatic prerequisite for task completion; it is not a psychological preference. The Machine Payments Protocol, co-authored by Stripe and Tempo Labs, formalizes this reality by reviving the HTTP 402 “Payment Required” status code into a standardized, machine-readable challenge-response framework.

This week’s reports

1️⃣What actually makes launching a card programme hard?

2️⃣What if the future of digital money is not stablecoins, but deposits in a new form?

3️⃣What happens when money becomes programmable but accountability does not scale with it?

4️⃣Impact of fintech in Australia

5️⃣How crypto fraud strategies are evolving for 2026

6️⃣What if tokenisation is not about efficiency, but about rebuilding the financial system from first principles?

7️⃣Revolut’s numbers are extreme

This week’s insights

1️⃣How Institutional Tokenization Is Actually Structured Today

2️⃣Post-Brex Ramp vs Mercury

3️⃣What is Required to Establish and Manage a BIN Sponsorship Program?

4️⃣What is actually driving fintech adoption in the Middle East right now?

5️⃣What is Ramp actually building?

6️⃣The 3% Problem: How AI Spots Payment Anomalies Before Revenue Slips Away

7️⃣Why Developers Should Care About Stablecoins

How Institutional Tokenization Is Actually Structured Today

Most people still talk about tokenization as a technology layer. That framing is already outdated.

What this landscape makes clear is that tokenization has become a full institutional stack.

-> You now have asset managers like BlackRock, Franklin Templeton, and Apollo actively issuing tokenized products.

-> You have a dedicated issuance layer with platforms like Securitize, Tokeny, and Centrifuge structuring assets into programmable instruments.

-> You have custody and wallet infrastructure from players like Fireblocks, Coinbase Custody, and BNY Mellon securing those assets.

-> You have trading venues, both crypto-native like Uniswap and Binance, and traditional infrastructure like Nasdaq and Archax, converging into the same market structure.

And then you have settlement layers where this becomes real infrastructure. JPMorgan Kinexys, Circle, SWIFT, CLS, Fnality. It seems like companies a re integrating into the core of financial infrastructure.

___

The key shift is this:

Read more

The financial infrastructure of the internet is being re-architected. On March 18, 2026, the launch of the Machine Payments Protocol (MPP) and the Tempo mainnet signaled the possible end of the human-centric checkout era. Actually, this could be a fundamental shift in how value moves across the web. For thirty years, the internet lacked a native value layer, forcing the industry to build increasingly complex user interfaces around legacy credit card rails. These interfaces, designed for human psychology, have become the primary bottleneck for the emerging agentic economy.

I believe the fundamental friction of the modern web is the “human-in-the-loop” requirement for commerce. AI agents do not use browsers. They do not click buttons, they do not solve CAPTCHAs, and they do not possess the cognitive friction that leads to cart abandonment. When an AI agent requires a resource, payment is a programmatic prerequisite for task completion; it is not a psychological preference. The Machine Payments Protocol, co-authored by Stripe and Tempo Labs, formalizes this reality by reviving the HTTP 402 “Payment Required” status code into a standardized, machine-readable challenge-response framework.

The failure of human-centric micropayments

There are three generations of failure in the micropayments graveyard. Every previous attempt failed because the assumed payer was a person. Humans calculate mental transaction costs for even the smallest payments. This leads to a hesitation that breaks the economics of digital content and services.

MPP succeeds where these failed because it eliminates the human decision point. For an agent, the cost of a transaction is a variable in an optimization function; it does not rely on an emotional event. This allows for sub-cent transaction models to finally become viable at scale.

MPP technical architecture

The core of MPP is its reliance on standard HTTP semantics, specifically the “Payment” HTTP Authentication Scheme proposed to the IETF. The protocol is a deterministic state machine that operates in a single round trip, removing the need for API keys or pre-existing accounts.

The five-step request-response cycle

The protocol operates through a formal sequence that embeds payment directly into the transport layer:

1. Resource Request: The agent initiates a standard request to a protected endpoint, such as POST /api/search.

2. Payment Challenge: The server intercepts the request and returns an HTTP 402 Payment Required status. The response includes a WWW-Authenticate: Payment header containing the payment challenge, which specifies the price, accepted currencies, and recipient address.

Read more

We dive into why Polygon is doubling down on payments, how stablecoins and tokenized deposits are reshaping money move…

Read more

Video of the Week

Deep Dive of the Week

The hidden liability of agentic commerce

In the last decade, we have watched the global financial system transition from physical cards to digital wallets and from manual bank wires to real-time rails. Each of these shifts was marketed as a convenience play for the consumer. In reality, each shift was an architectural reset for the merchant. We are now entering the most volatile phase of this evolution: agentic commerce. This is not just another checkout button. It is a structural rewiring of how demand is captured, how intent is verified, and how liability is distributed. I define agentic commerce as the transition from a world where humans click “buy” on a screen to a world where autonomous software agents initiate, authorize, and settle transactions on behalf of users.

For the fintech founders, CEOs, and payments strategists reading this, the promise of a frictionless $1 trillion market is enticing. But I am here to tell you that the current payment infrastructure is fundamentally broken for this new reality. The industry is currently ignoring a massive hidden liability. When an AI agent buys the wrong product or acts without explicit consent, the merchant is the one left holding the bill. Traditional fraud detection is dead in this environment. If your risk model still relies on monitoring typing cadence or mouse movements, you are already obsolete. We are moving into a “fifth participant” model where the AI agent acts as a new layer of intermediation between the cardholder and the merchant. This fifth actor breaks the historical four-party model of liability and pushes risk downstream to you.

This week’s reports

1️⃣How Retail Banks Can Put AI Agents to Work

2️⃣How crypto cards quietly became the main bridge between stablecoins and real-world spending

3️⃣Blockchain, Graph Theory, and the Path to Streaming Financial Services

4️⃣From AI-Assisted Banking to Autonomous Banks

5️⃣Agentic AI will compress margins and reshape financial infrastructure

6️⃣Next-Gen Acquiring Processing

7️⃣Deep Dive into Fintech Mercury

This week’s insights

1️⃣Not All Rails or Stablecoins are Equal

2️⃣Fiat-to-crypto infrastructure is not a payments problem. It is a ledger problem

3️⃣Digital Commerce Payment Platforms

4️⃣Understanding custody models for tokenized assets

5️⃣Agentic Commerce as a Proportion of Total E-commerce

6️⃣One of the fastest-growing financial institutions globally

7️⃣4 Payment KPIs That Reveal Where Revenue Is Actually Leaking

Not All Rails or Stablecoins are Equal

The “average of averages” is often misleading, especially in cross-border payments.

Stablecoins are typically fast (assuming no fiat on-/off-ramp dependencies), but not always cost-effective, with highly volatile pricing. Data from Token Terminal shows Ethereum’s average daily transaction fee over three years was $0.134, with peaks up to $29.58. The median was $0.0074, peaking at $13.093. The variability is significant.

Layer-2 stablecoins offer a different profile, with lower fees despite being built on layer-1s like Ethereum. Over the same period, Arbitrum One recorded an average fee of $0.0127, peaking at $0.7033, with a median of $0.0034 and peak of $0.51072.

The data shows Ethereum’s dominance is driven less by performance and more by liquidity and distribution. Importantly, these figures reflect only on-chain “gas” fees—composed of base and priority fees.

In reality, costs extend across the full value chain: on-/off-ramp fees (including AML/KYC and FX), liquidity provider fees, and PSP or MTO charges tied to transaction size. Additional fees may apply on payout. Users can reduce costs by transacting directly via wallets.

Read more

In the last decade, we have watched the global financial system transition from physical cards to digital wallets and from manual bank wires to real-time rails. Each of these shifts was marketed as a convenience play for the consumer. In reality, each shift was an architectural reset for the merchant. We are now entering the most volatile phase of this evolution: agentic commerce. This is not just another checkout button. It is a structural rewiring of how demand is captured, how intent is verified, and how liability is distributed. I define agentic commerce as the transition from a world where humans click “buy” on a screen to a world where autonomous software agents initiate, authorize, and settle transactions on behalf of users.

For the fintech founders, CEOs, and payments strategists reading this, the promise of a frictionless $1 trillion market is enticing. But I am here to tell you that the current payment infrastructure is fundamentally broken for this new reality. The industry is currently ignoring a massive hidden liability. When an AI agent buys the wrong product or acts without explicit consent, the merchant is the one left holding the bill. Traditional fraud detection is dead in this environment. If your risk model still relies on monitoring typing cadence or mouse movements, you are already obsolete. We are moving into a “fifth participant” model where the AI agent acts as a new layer of intermediation between the cardholder and the merchant. This fifth actor breaks the historical four-party model of liability and pushes risk downstream to you.

The structural collapse of human-centric fraud detection

The existing payment architectures are built on the assumption of human presence. This is a foundational flaw in the age of automation. Traditional systems rely on behavioral biometrics and device fingerprinting to establish trust. In an agentic world, these signals disappear entirely. There is no typing cadence to measure when a script is making the call. There is no mouse movement when a server-side agent is interacting with an API. This creates a total loss of the traditional fraud signal.

I see this as a “shifting trust perimeter.” Payments no longer originate on your storefront, where you control the environment. They originate in external LLMs, personal assistants, and procurement bots. This decentralization of the checkout experience means you can no longer rely on your own telemetry to verify the buyer. You are now forced to trust a software agent you did not build and cannot inspect. This creates several structural failures that I will break down in this report.

The disappearance of these signals will lead to a spike in “friendly fraud” and “hallucination disputes.” If an agent orders the wrong item because of a misinterpretation of user intent, the merchant is likely to be held liable for the return and the chargeback. Liability follows the Merchant of Record. In every protocol currently being proposed, from OpenAI to Google, the merchant remains the MoR. You are taking on all the risk of autonomous execution with none of the historical protections.

The protocol landscape

The industry is currently in a “protocol war” to define how agents and merchants communicate. I have analyzed the major standards emerging in 2025 and 2026. Each protocol addresses a different layer of the stack, from the application layer to the settlement layer. I think it is a mistake to view these as competing products. They are building a “capability ladder” that determines how much risk a merchant is prepared to absorb.

Agentic commerce protocol (ACP)

OpenAI and Stripe co-developed ACP as an open standard (Apache 2.0) to make checkouts “agent-ready”. ACP is the application layer for conversational commerce. It is optimized for speed and convenience within chat interfaces like ChatGPT.

The core technical primitive of ACP is the Shared Payment Token (SPT). Instead of exposing raw card data to the agent, the system generates a scoped, ephemeral token bound to a specific merchant and a specific amount. This protects the consumer’s credentials but does little to protect the merchant from intent disputes. In a typical ACP flow, the agent manages the checkout interface, while the merchant manages the data model and fulfillment.

ACP is ideal for high-intent, human-in-the-loop purchases where a user is actively chatting with an assistant. However, it lacks the robust governance needed for fully autonomous, background purchases. This is where Google’s AP2 comes in.

Agent payments protocol (AP2)

Google led a coalition of 60 partners to build AP2, focusing on the authorization and traceability layer. I consider AP2 to be the governance layer for agentic commerce. It introduces “mandates,” which are cryptographically signed digital contracts that define the agent’s boundaries.

These mandates serve as verifiable proof of user instructions. If an agent tries to spend $500 when the user only authorized $300, the AP2 layer rejects the transaction before it ever reaches the processor. This provides the audit trail merchants need to defend against chargebacks.

I have broken down the AP2 mandate types to illustrate their roles in the trust chain:

• Intent Mandate: A high-level description of what the user wants (e.g., “Buy salmon and olive oil from Example Wholesale”).

• Payment Mandate: A specific authorization bound to a cart ID and a final amount.

• Payment Receipt: Closes the audit loop with a merchant confirmation ID, providing non-repudiatable proof of execution.

AP2 is rail-agnostic. It can carry instructions for cards, bank transfers, or crypto assets through its x402 extension. For the institutional reader, AP2 represents the most serious attempt to create a standard for compliance and multi-party governance in autonomous commerce.

Universal commerce protocol (UCP)

UCP is Google’s broader attempt to standardize the entire commerce lifecycle. While ACP and AP2 focus on the transaction, UCP handles the context. It enables “native checkout” inside AI Mode in Search and Gemini, keeping the merchant as the seller of record while the transaction happens on a Google surface.

I think UCP’s most significant contribution is the standardization of discovery and post-purchase flows. It defines JSON-RPC 2.0 schemas for adding items to carts, calculating logistics, and managing returns. This allows a merchant to “build once” and distribute their catalog to any UCP-compatible agent.

The identity crisis: differentiating agents from bots

The biggest operational challenge for merchants is the “identity gap.” If you cannot differentiate between a legitimate AI agent and a malicious scraper, you have to block all automated traffic to protect your infrastructure. This is a revenue killer in an agentic world. We are seeing two distinct approaches to solving this through the major card networks.

Visa Trusted Agent Protocol (TAP)

Visa’s TAP provides a cryptographic standard for agents to prove their identity and authorization directly to the merchant. The trust model here is robust because it is built on three distinct signatures:

1. Agent Recognition Signature: Sent in the HTTP header to ensure the agent is Visa-approved.

2. Consumer/Device Identity: A signed object in the request body that allows the merchant to check for existing loyalty accounts or prior interactions.

3. Payment Container Signature: A hash of the payment credentials to ensure they were not tampered with during the handoff.

TAP allows agents to pass “context” without requiring a full login. This is the solution to the “cold start” problem in e-commerce. An agent can prove it is acting for a known user and provide their shipping preferences and loyalty ID, allowing for a personalized guest checkout.

Mastercard Agent Pay acceptance framework

Mastercard’s framework is designed for mass-market scalability. It leverages the “Web Bot Auth” standard (IETF RFC 9421) at the CDN layer. This is a “no-code” approach for merchants. By implementing this at the edge, you can verify agent authenticity and block untrusted traffic without deploying new code to your backend.

I think this is the most pragmatic path for the average merchant. It allows them to accept “agentic tokens” formatted as standard card fields through their existing checkout forms. This preserves backward compatibility with the existing card rails while adding a layer of cryptographic verification.

The tokenization tier: universal vs. network tokens

I have long argued that tokenization is the only way to secure the payments lifecycle. In agentic commerce, tokenization becomes programmable infrastructure. It is no longer just about PCI compliance; it is about “replayability” and “auditability”.

There is a clear battle between three families of tokens. For the enterprise leader, the choice of token determines your level of vendor lock-in and your ability to route across multiple acquirers.

1. PSP-Specific Tokens: Minted by a single gateway. These are “hostage data” playbooks. They offer no portability and create total dependency on a single provider.

2. Network Tokens: Issued by schemes like Visa and Mastercard. They provide issuer-level trust and include lifecycle updates (e.g., when a card is refreshed). They improve authorization rates and lower fraud.

3. Universal Tokens: Minted and stored by an independent vault (e.g., IXOPAY). These decouple tokenization from processing. They allow you to “tokenize once, route anywhere”.

I advocate for a “dual-tier” approach. Merchants should own their universal tokens as the central “enterprise handle.” These tokens should resolve to network tokens at the edge where supported. This captures the performance benefits of the networks while maintaining the sovereignty of the merchant’s data.

In an agentic flow, tokenization expands to include intent metadata. I think it is essential to tokenize the “mandate” alongside the payment credential. This pairing creates a full machine-verifiable record of the transaction. If the user later disputes the charge, you have the universal token (proving identity) linked to the intent token (proving authorization). This is your only defense in the coming chargeback wave.

The settlement revolution: x402 and machine-native money

I believe that traditional credit card rails are fundamentally incompatible with the machine economy. High transaction fees (2.9% + $0.30) and T+2 settlement times are too slow and expensive for autonomous agents. If an AI agent needs to pay $0.05 for a single API call, it cannot use a credit card.

The x402 protocol, developed by Coinbase, is the first real settlement layer for agents. It revives the HTTP 402 “Payment Required” status code to enable instant stablecoin payments directly over the web protocol.

The x402 V2 architecture is particularly compelling for machine-to-machine commerce. It uses a “Challenge/Commitment” model:

• Challenge: The server responds with a 402 status code and a structured quote (amount, asset, blockchain, destination, and a one-time nonce).

• Commitment: The agent’s wallet signs a payment authorization specific to that quote and retries the request with a X-PAYMENT header.

• Settlement: The merchant verifies the signature and completes the trade instantly.

I think x402 is the “Stripe for AI agents”. It allows for micro-transactions with zero processing fees beyond on-chain gas. While ACP and UCP handle the discovery of physical goods, x402 is optimized for software paying for software APIs, data feeds, and compute resources.

Phase zero: architecting for scale

Most internal roadmaps of early adopters are failing because they are treating agentic commerce as a “feature” rather than an “operating model shift”. I advocate for a “Phase Zero” approach that builds the foundational primitives before the volume arrives.

Design principles for the agentic stack

Five core design principles must be implemented in the next 12 months to avoid architectural obsolescence.

1. Modular Architecture: Avoid monolithic structures. Build the payment stack as a set of independent, interchangeable parts for intent capture, policy enforcement, and routing.

2. Automation First: Scaling requires systems that run themselves. If a task is performed more than twice, it must be automated to prevent human bottlenecks as agent traffic scales from 1% to 10%.

3. Data-Driven Telemetry: You need robust telemetry from day one. You must be able to track “agent drift” when an agent’s behavior deviates from its authorized intent.

4. Elasticity: Systems must handle variable loads. Agentic traffic is not predictable; it can spike 10x in a single session.

5. Documentation as Code: Institutional knowledge is the enemy of scale. All agentic configurations and rules must be machine-readable and centralized.

The implementation roadmap

The move to agentic commerce happens in four distinct phases of infrastructure maturity.

The merchant value and the competitive moat

Many payment leaders are hesitant to invest because the volume is not yet there. I think this is a failure of vision. Agentic commerce is not just a new channel; it is a new way to reduce friction and capture demand closer to the moment of intent.

Merchants who prepare their trust infrastructure today will gain three strategic advantages:

• Reduced Liability: Comprehensive audit trails and intent tokenization will drastically lower the cost of chargeback disputes.

• Innovation Velocity: A protocol-agnostic stack allows you to adopt new agentic standards (like Google’s UCP or OpenAI’s ACP) without re-platforming.

• Operational Efficiency: Moving from manual, human-driven checkouts to machine-verifiable records lowers the overhead of fraud detection and customer support.

I believe that trust is the only durable competitive moat in an autonomous world. If an agent can trust that your storefront is “machine-readable” and “intent-safe,” it will prefer your site over a competitor’s legacy interface.

The role of trust orchestration

I will be blunt: the role of the payment orchestrator has expanded. They must now orchestrate trust, not just transactions. This requires moving orchestration “above” individual gateways and protocols.

I think the critical functions of a trust orchestrator include:

• Aggregating Trust Signals: Individually tokenizing identity, consent, and behavior into a single “universal token” that travels across the payment journey.

• Interoperability: Providing a single integration that supports every agentic protocol—Google, Amazon, OpenAI—without building separate pipes.

• Intelligent Monitoring: Continuous visibility into payment health and authorization patterns to detect rogue agents in real time.

My take: control is the new currency

I have spent this report detailing the technical and operational risks of agentic commerce. But the throughline is control. Whether it is Google and Stripe fighting over protocols, or Visa and Mastercard fighting over identity, the battle is about who owns the infrastructure layer of the machine economy.

For the merchant, control means owning your tokens and your audit trails. If you delegate your trust to a third-party protocol, you are delegating your revenue. I think the companies that win will not just process transactions; they will define the rails the rest of the ecosystem runs on.

Your existing payments stack is a liability. It was built for humans. The transition to agentic commerce is the biggest unlock of the decade, but it is also the biggest threat to those who rely on the behavioral moats of the past. The winners will be those who architect a neutral trust layer today. Everything else is just legacy infrastructure waiting to be optimized away.

Sources

1. WC_Gated_Education_The_Merchant_Risk_Problem_in_Agentic_Commerce_WP.pdf

2. Agentic Commerce & Merchant Risk | Building the New Trust Layer, accessed March 22, 2026, https://www.ixopay.com/whitepapers/the-merchant-risk-problem-in-agentic-commerce

3. Agentic Commerce Is Coming: What Payment Leaders Need to ..., accessed March 22, 2026, https://www.ixopay.com/blog/agentic-commerce-is-coming-what-payment-leaders-need-to-build-now

4. Agentic Commerce Protocol (ACP), accessed March 22, 2026,

https://www.agenticcommerce.dev/

1. Google Universal Commerce Protocol (UCP) Guide, accessed March 22, 2026, https://developers.google.com/merchant/ucp

2. Agentic Commerce Protocol - OpenAI Developers, accessed March 22, 2026, https://developers.openai.com/commerce/guides/get-started

3. What the ACP vs AP2 agentic payments comparison means for you - Grid Dynamics, accessed March 22, 2026, https://www.griddynamics.com/blog/agentic-payments

4. ACP vs. AP2 vs. TAP: The Protocol Wars of Agentic Commerce | PayRam, accessed March 22, 2026, https://payram.com/blog/acp-vs-ap2-vs-tap

5. Integrate the Agentic Commerce Protocol - Stripe Documentation, accessed March 22, 2026, https://docs.stripe.com/agentic-commerce/protocol

6. Announcing Agent Payments Protocol (AP2) | Google Cloud Blog, accessed March 22, 2026, https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol

7. Developer’s Guide to AI Agent Protocols, accessed March 22, 2026, https://developers.googleblog.com/developers-guide-to-ai-agent-protocols/

8. Agentic Payments Explained: ACP, AP2, and x402 | Orium, accessed March 22, 2026, https://orium.com/blog/agentic-payments-acp-ap2-x402

9. About the Universal Commerce Protocol (UCP) and UCP-powered checkout feature on Google - Google Merchant Center Help, accessed March 22, 2026, https://support.google.com/merchants/answer/16837055?hl=en-IE

10. About the Universal Commerce Protocol (UCP) and UCP-powered checkout feature on Google - Google Merchant Center Help, accessed March 22, 2026, https://support.google.com/merchants/answer/16837055?hl=en

11. Building the Agentic Commerce Stack: How to Connect the x402 Payment Protocol with the Universal… - Medium, accessed March 22, 2026, https://medium.com/@iamanuragsaini/building-the-agentic-commerce-stack-how-to-connect-the-x402-payment-protocol-with-the-universal-ac7241974f08

12. UCP vs ACP: Which Agentic Commerce Protocol Should Retailers Choose? - Paz.ai, accessed March 22, 2026, https://www.paz.ai/blog/ucp-vs-acp-which-agentic-commerce-protocol-should-retailers-choose

13. Trusted Agent Protocol - Visa Developer, accessed March 22, 2026, https://developer.visa.com/use-cases/trusted-agent-protocol

14. GitHub - visa/trusted-agent-protocol: Establishing a universal standard of trust between AI agents and merchants for the next phase of agentic commerce., accessed March 22, 2026, https://github.com/visa/trusted-agent-protocol

15. Specifications - Trusted Agent Protocol - Visa Developer, accessed March 22, 2026, https://developer.visa.com/capabilities/trusted-agent-protocol/trusted-agent-protocol-specifications

16. Trusted Agent Protocol - Visa Developer, accessed March 22, 2026, https://developer.visa.com/capabilities/trusted-agent-protocol

17. Agentic token framework: Driving trusted AI transactions - Mastercard, accessed March 22, 2026, https://www.mastercard.com/global/en/news-and-trends/stories/2025/agentic-commerce-framework.html

18. Visa and Mastercard both launch new agentic AI payments tools - Digital Commerce 360, accessed March 22, 2026, https://www.digitalcommerce360.com/2025/10/16/visa-mastercard-both-launch-agentic-ai-payments-tools/

19. Getting Started - Trusted Agent Protocol - Visa Developer, accessed March 22, 2026, https://developer.visa.com/capabilities/trusted-agent-protocol/docs-getting-started

20. Agentic Payments: x402 and AI Agents in the AI Economy - Galaxy, accessed March 22, 2026, https://www.galaxy.com/insights/research/x402-ai-agents-crypto-payments

21. Agentic payments protocols compared: Which is best for your AI agents? (ACP vs AP2 vs x402) - Crossmint, accessed March 22, 2026, https://www.crossmint.com/learn/agentic-payments-protocols-compared

22. The x402 protocol and agentic commerce - The Paypers, accessed March 22, 2026, https://thepaypers.com/payments/expert-views/x402-standardising-the-protocol-for-agent-to-agent-commerce

23. Sam Boboev | Substack - Fintech Wrap Up, accessed March 22, 2026, https://www.fintechwrapup.com/t/deep-dive

24. Sam Boboev (@samboboev): “ Deep Dive: How Fintech Block Is Replacing Processes and People with Agents Block just made the cleanest public claim so far that “AI gains” can justify a headcount reset at scale. Jack Dorsey framed the February 2026 reduction as an “AI overhaul,” cutting more than 4,000 rol…” - Substack, accessed March 22, 2026,

1. accessed March 22, 2026, uploaded:ffggghhj.jpg-6feb5b1b-b147-489e-8bbb-4151b9205c5b

2. Webinar | Architecting Agentic Payments: The Two Pillars of Success, accessed March 22, 2026, https://www.ixopay.com/webinars/webinar-replay-architecting-agentic-payments

3. Introducing the Agentic Commerce Suite: A complete solution for selling on AI agents, accessed March 22, 2026, https://stripe.com/blog/agentic-commerce-suite

4. Sam Boboev (@samboboev): “Welcome to the new edition of the Fintech Wrap Up. This week’s themes orbit around infrastructure power shifts in fintech — from who controls payment rails to how capital markets are repricing platforms and digital assets. What happened? Fintech’s biggest players continue to sc…” - Substack, accessed March 22, 2026,

1. Sam Boboev (@samboboev): “Welcome to the new edition of the Fintech Wrap Up — where things are getting a bit more autonomous. What happened? This week’s theme is clear: finance is being rewired for a world where software, not humans, is increasingly in the driver’s seat. Agentic commerce is emerging as…” - Substack, accessed March 22, 2026,

Disclaimer:

Fintech Wrap Up aggregates publicly available information for informational purposes only. Portions of the content may be reproduced verbatim from the original source, and full credit is provided with a “Source: [Name]” attribution. All copyrights and trademarks remain the property of their respective owners. Fintech Wrap Up does not guarantee the accuracy, completeness, or reliability of the aggregated content; these are the responsibility of the original source providers. Links to the original sources may not always be included. For questions or concerns, please contact us at sam.boboev@fintechwrapup.com.

We go beyond the usual narrative and unpack what is actually broken: fragmented infrastructure, compliance complexity, local rails, and the reality that there is no one-size-fit…

Read more

Video of the Week

Deep Dive of the Week

Agentic AI in Financial Crime Fighting

The global financial system is facing a structural crisis. In 2025, illicit financial activity surged to an estimated $4.4 trillion, a figure that represents a $1.3 trillion increase over just two years. This 19.2% compound annual growth rate indicates that financial crime is not merely increasing; it is industrializing. This surge means legacy compliance models are failing. Traditional rule-based systems and manual human review are mathematically incapable of keeping pace with adversarial AI.

The industry currently operates in a state of high-cost inefficiency. Banks commonly allocate 10% to 15% of their total headcount to Know Your Customer (KYC) and Anti-Money Laundering (AML) activities, yet they detect only about 2% of global financial crime flows. This delta between operational spend and effectiveness is the “compliance trap.” I believe agentic AI is the only credible exit strategy for this trap.

Agentic AI represents a shift from “assistive” technology to “autonomous” execution. While generative AI (GenAI) summarizes data and analytical AI identifies patterns, agentic AI has the capacity to plan, execute, and adapt sequences of actions to reach a specific objective. It is the difference between a chatbot that writes a summary and a digital worker that investigates a case.

This week’s reports

1️⃣Stablecoins Are Becoming a New Settlement Rail in Global Payments

2️⃣Agentic Commerce Is Changing the Structure of Retail

3️⃣Stablecoin Payments: The Volume Illusion

4️⃣Tokenization’s Real Bottleneck: Interoperability in Digital Capital Markets

5️⃣Crypto taxation is no longer fragmented experimentation

6️⃣Stablecoins are becoming domestic payment infrastructure

7️⃣Tokenized stocks transition layer

This week’s insights

1️⃣The Tokenised Money Framework

2️⃣When AI agents start shopping, who’s responsible for the chargeback?

3️⃣European Banks Prepare for Tokenized Finance

4️⃣Neobank Market Map by Artemis

5️⃣Big crypto is coming for big banking

6️⃣The stablecoin infrastructure framework

7️⃣How Tokenization Is Evolving for Autonomous Payments

The Tokenised Money Framework

As monetary instruments move onto distributed ledgers, multiple classification attempts have emerged. No widely accepted taxonomy exists. Different actors define these instruments through different lenses:

- Policymakers classify them through regulatory frameworks and legal claims.

- Financial institutions focus on functional form deposit tokens, tokenised deposits, programmability, and settlement use cases.

- Crypto-native actors classify them by reserve structures and market adoption patterns.

This fragmentation complicates comparisons across instruments and weakens analysis of their performance against the three functions of money: unit of account, medium of exchange, and store of value. It also complicates regulatory discussions.

A two-layer taxonomy addresses this gap. The framework defines instruments based on the underlying claim and maps the design features that determine functionality.

Tokenised money refers to DLT-based claims referencing fiat money used for payments or value storage. The definition also includes quasi-monetary claims that perform monetary functions under certain conditions and may represent proportional claims on underlying assets rather than guaranteed par redemption.

____

The first layer classifies instruments across four dimensions:

Read more

The global financial system is facing a structural crisis. In 2025, illicit financial activity surged to an estimated $4.4 trillion, a figure that represents a $1.3 trillion increase over just two years. This 19.2% compound annual growth rate indicates that financial crime is not merely increasing; it is industrializing. This surge means legacy compliance models are failing. Traditional rule-based systems and manual human review are mathematically incapable of keeping pace with adversarial AI.

The industry currently operates in a state of high-cost inefficiency. Banks commonly allocate 10% to 15% of their total headcount to Know Your Customer (KYC) and Anti-Money Laundering (AML) activities, yet they detect only about 2% of global financial crime flows. This delta between operational spend and effectiveness is the “compliance trap.” I believe agentic AI is the only credible exit strategy for this trap.

Agentic AI represents a shift from “assistive” technology to “autonomous” execution. While generative AI (GenAI) summarizes data and analytical AI identifies patterns, agentic AI has the capacity to plan, execute, and adapt sequences of actions to reach a specific objective. It is the difference between a chatbot that writes a summary and a digital worker that investigates a case.

The $4.4 trillion market of illicit activity

The scale of global financial crime in 2026 is staggering. Criminal networks now operate with the coordination and scale of multinational corporations. Fraud scams are growing at 19.3%, nearly twice the rate of traditional bank fraud. This growth is powered by AI-driven attacks that overwhelm static defenses.

Global Financial Crime Landscape 2025-2026

90% of financial crime professionals report a measurable increase in AI-driven attacks. These attacks use automation to create synthetic identities, execute deepfake-driven scams, and coordinate cross-channel fraud campaigns. I view the current state of financial crime as a “scamdemic” that necessitates a move from reactive policing to predictive defense.

The obsolescence of no-code and rule-based frameworks

For a decade, “no-code” was the benchmark for risk operations. It allowed compliance teams to build rules without engineering support. However, as crime volumes escalated, the analyst became the bottleneck. In traditional AML, up to 95% of alerts are false positives. Building a single Suspicious Activity Report (SAR) can take four or more days.

No-code tooling is no longer sufficient. The requirement is now for AI Risk Infrastructure. This infrastructure executes the full financial crime lifecycle: detecting risk in real time, investigating alerts end-to-end, and producing regulator-ready filings. Unit21’s 2026 relaunch signals this transition. Their platform moved from being a no-code rules engine to an agentic system where AI agents tune detection logic and conduct investigations without human analysts driving every step.

Defining agentic AI in risk operations

Agentic AI refers to systems that act with a degree of autonomy toward defined goals. In financial crime fighting, this means the AI can decide which data sources to query, how to interpret inconsistent information, and when to escalate a case.

Comparison of AI Generations in Compliance

The productivity potential of agentic AI is a 20-fold increase over manual practitioners. I categorize these agents into squads that mirror human roles along the value chain. Retrieval-Augmented Generation (RAG) agents handle data extraction from profit-and-loss statements and beneficial ownership documents. Data pipeline agents orchestrate ETL processes and perform entity resolution across fragmented datasets. Research agents monitor market trends and counterparty patterns, while validation agents review agent outputs to ensure quality.

Context engineering vs prompt engineering

The hardest engineering challenge in agentic AI is not writing better prompts; it is context engineering. To produce an auditable investigation narrative, the model must receive the exact right evidence without overloading its context window. LLMs are based on transformer architecture, where every token attends to every other token, resulting in n2 relationships. This leads to attention scarcity as context length increases.

Effective context engineering is the science of curating high-signal tokens to maximize the likelihood of a desired outcome. For example, Unit21 leverages their rich dataset from 7 years of human reviews to work out the optimal context required to complete given tasks. These tasks are then evaluated against historical human investigations, completed by high performing analysts, to ensure correctness, consistency, and effectiveness.

Evaluation is performed using “LLM-as-a-judge” architectures. A secondary, more capable model evaluates the quality of the primary agent’s output, creating a self-checking layer that flags inconsistencies before they reach a human reviewer. This is supplemented by citation validation, where the system verifies that agent claims are grounded in retrieved data rather than model inference.

The AI investigation workflow

When an alert enters the queue, the AI Investigation Agent follows a structured workflow rather than starting from a blank page.

1. Signal gathering: The agent retrieves the transaction history, entity profile, risk scores, and watchlist matches. It navigates across disparate screens to assemble the context a senior analyst would require.

2. Workflow orchestration: The agent follows modular steps configured to the institution’s standard operating procedures (SOPs). This includes checking prior alert history, running OSINT searches, and cross-referencing sanctions lists.

3. Findings assembly: The agent produces a structured package containing a written narrative, evidence logs, and a recommended disposition. The reasoning is explicit and traceable.

The “human-in-the-loop” model remains the default for final dispositions. Analysts approve, modify, or override the agent’s package, ensuring human accountability.

The rise of agentic attacks

Financial crime is accelerating because the attackers are already agentic. In 2026, we are watching seven specific AI-driven fraud vectors and agentic attacks.

1. Polymorphic phishing agents

These agents do not send mass emails. They embed themselves in compromised inboxes, analyze historical threads, and insert themselves into high-trust conversations. They mirror the victim’s typing rhythms and working hours to bypass reputation-based security systems.

2. Invoice-timed malware

Malware families like Predator map business relationships and invoice schedules. They send counterfeit invoices a day before the legitimate ones are expected, targeting operational psychology where speed often overrides scrutiny.

3. Deepfake-as-a-Service (DFaaS)

Attackers use real-time APIs for voice cloning and video manipulation to bypass biometric liveness checks. Deepfake fraud jumped by 1100% in the US and 3400% in Canada between 2025 and 2026.

4. Synthetic identity farms

AI agents manage synthetic profiles, programmatically building credit history over 18 months. These identities remain dormant until a coordinated “activation event” triggers them to drain balances across multiple institutions simultaneously.

5. Ghost touch attacks

Remote Access Trojans combined with electromagnetic interference (EMI) hardware allow attackers to inject touch inputs and install malware without physical interaction.

6. Automated CVE weaponization

Agents ingest public vulnerability disclosures and generate functional exploit code at machine speed, attacking systems before they can be patched.

7. Automated chain-hopping

Laundering agents orchestrate cross-chain strategies, fragmenting stolen funds into tens of thousands of micro-transactions under $10. This makes the economic cost of manual tracing higher than the value of the assets.

Operational gains and case studies

The results of agentic AI deployment are measurable and significant. For example, according to Unit21, Underdog Fantasy reduced its alert volume by 72% using agentic AI, and Nexo reduced false positives by 57% and is targeting 80% after deploying Unit21’s AI agents.

Impact of Agentic AI on Operations

In transaction monitoring, agentic AI reduces noise by ranking cases by expected risk and using investigator copilots to summarize context. Feedback loops from analyst decisions allow the models to adapt as typologies change. I find that AI Rule Recommendations are critical here, as they analyze patterns across historical dispositions to suggest threshold adjustments that teams “didn’t know they needed”.

The three failure modes of AI agents

Most early deployments of AI agents fail because of poor guardrails rather than weak models.

1. The hallucinating investigator: This happens when teams provide too much context and open-ended prompts. In adversarial environments, the model fills data gaps with plausible but incorrect narratives. The solution is to use “atomic agents” with narrow decision boundaries.

2. The over-suspicious agent: Pattern-driven training without contextual grounding leads to over-escalation. For example, flagging high-value payments between related internal accounts as “layering”. Grounding questions must be injected into the agent’s logic to prevent default conclusions of fraud.

3. The black box agent: Producing conclusions that are not defensible to regulators. Accurate outputs without a chain of evidence are a liability. Agents must pull data deterministically and focus on structured documentation.

Regulatory environment in 2026

The regulatory stance on agentic AI is shifting from skepticism to expectation. On March 6, 2026, the White House released its National Cyber Strategy and an Executive Order focused on “Combating Cybercrime, Fraud, and Predatory Schemes”. This strategy encourages the use of AI-powered cyber-defense to better protect government networks and scale network defense.

FINRA’s 2026 oversight report includes a dedicated AI section, emphasizing that firms must be able to explain how AI is used and how outputs are tested. Agentic AI raises the stakes because these systems take actions, not just generate content. Transparency into assumptions and accountability for outcomes are now non-negotiable.

2026 Regulatory Milestones

The Financial Action Task Force (FATF) has approved new strategic publications on cyber-enabled fraud and virtual asset risks, encouraging supervisors to leverage innovative technology to return stolen funds to victims. In Europe, banks are moving from static, periodic reviews to “continuous customer understanding” under the scrutiny of the new AML watchdog, AMLA.

Network-level defense

The broader intelligence layer in agentic AI comes from the consortium model. Anonymized, aggregated signal across a customer base allows for the identification of emerging fraud patterns before any single institution has seen enough volume to detect them independently. This network effect allows agentic systems to suggest logic that counters threats across the entire financial ecosystem.

I believe that fragmented systems for onboarding, monitoring, and payment screening create gaps that criminals exploit. A unified screening suite that applies the same logic consistently across the full customer lifecycle is essential. Transparency and explainability build trust and ensure decisions remain aligned with both internal policies and external regulations.

Predictive defense and digital workers

As we move through 2026, the distinction between private stablecoins and public digital money is becoming a critical strategic consideration. The fusion of fraud and AML operations is not just operational convergence; it is a deeper integration of the technology stack.

Agentic AI systems are moving from the pilot phase to the core of AML defense. We are seeing a shift from simple pattern recognition to predictive systems that anticipate criminal activity before a transaction is even flagged. I’ll be blunt: legacy rule-based systems cannot keep up with the speed of instant payments.

The path to impact is driven by speed of adoption and a tailored operating model. Leading institutions are starting with pilot perimeters to prove impact before preparing for a full-scale rollout. Agentic AI is the next major innovation lever for KYC/AML, offering stronger compliance and a more streamlined customer experience.

I treat the adoption of agentic AI as a necessity for survival in the modern financial landscape. The $4.4 trillion in illicit activity is a reminder that the cost of inaction is too high. We must move from a workforce of manual executors to one of AI supervisors, managing a digital factory of agents that detect and investigate at machine speed.

Sources:

1. Nasdaq® Verafin Report Finds the Financial Crime Epidemic Reaching Alarming New Heights as Illicit Financial Activity Surges to $4.4 Trillion in 2025, accessed March 15, 2026, https://ir.nasdaq.com/news-releases/news-release-details/nasdaqr-verafin-report-finds-financial-crime-epidemic-reaching

2. 2026 Global Financial Crime Report - Nasdaq Verafin, accessed March 15, 2026, https://verafin.com/nasdaq-verafin-global-financial-crime-report/

3. NEWS: Financial crime activity reaches $4.4 trillion per year, growing nearly 20% annually, accessed March 15, 2026, https://www.amlintelligence.com/2026/03/news-financial-crime-activity-reaches-4-4-trillion-per-year-growing-nearly-20-annually/

4. how-agentic-ai-can-change-the-way-banks-fight-financial-crime.pdf

5. AI for Financial Crime: Real-Time Threats, Zero Lag - Sardine, accessed March 15, 2026, https://www.sardine.ai/blog/ai-for-financial-crime

6. Unit21 Relaunches as the Leader in AI Risk Infrastructure - Business Wire, accessed March 15, 2026, https://www.businesswire.com/news/home/20260310768649/en/Unit21-Relaunches-as-the-Leader-in-AI-Risk-Infrastructure

7. AI Agents for AML and Fraud: Why Unit21 Rebuilt Everything, accessed March 15, 2026, https://www.unit21.ai/blog/the-new-unit21-why-we-rebuilt-everything-around-ai-agents

8. AI-powered fraud: 5 trends financial institutions need to understand in 2026, accessed March 15, 2026, https://www.thomsonreuters.com/en-us/posts/corporates/ai-powered-fraud-5-trends/

9. The Future of AML Compliance: Strategic Predictions for 2026 - Feedzai, accessed March 15, 2026, https://www.feedzai.com/blog/future-aml-compliance-predictions/

10. Unit21 Named to the 2026 RegTech100 List for Agentic AI Innovation - Blog, accessed March 15, 2026, https://www.unit21.ai/blog/unit21-named-to-the-2026-regtech100-list-for-agentic-ai-innovation

11. Unit21: Agentic AI Platform for Fraud & AML Operations, accessed March 15, 2026, https://www.unit21.ai/

12. How AI Is Improving AML Software in 2026 | Alessa, accessed March 15, 2026, https://alessa.com/blog/how-ai-is-improving-aml-software-in-2026/

13. Unit21 Custom AI Agents: Next-Gen Fraud and AML Tools - Blog ..., accessed March 15, 2026, https://www.unit21.ai/blog/how-custom-ai-agents-are-transforming-fraud-and-aml-operations

14. How to Move Off Legacy Transaction Monitoring with AI | Unit21 - Blog, accessed March 15, 2026, https://www.unit21.ai/blog/ai-powered-transaction-monitoring-replacing-legacy-black-box-systems

15. Effective context engineering for AI agents - Anthropic, accessed March 15, 2026, https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents

16. Sardine’s blog, accessed March 15, 2026, https://www.sardine.ai/blog

17. AI-driven fraud vectors: 7 agentic attacks now live in 2026 - Sardine, accessed March 15, 2026, https://www.sardine.ai/blog/agentic-attacks

18. Synthetic Identity Document Fraud Surges 300% in the U.S. ..., accessed March 15, 2026, https://sumsub.com/newsroom/synthetic-identity-document-fraud-surges-300-in-the-u-s-sumsub-warns-e-commerce-healthtech-and-fintech-at-risk/

19. The 3 Failure Modes of Agentic AI in Financial Crime (and How to ..., accessed March 15, 2026, https://www.sardine.ai/blog/agentic-ai-financial-crime-failure-modes

20. White House’s New Cyber Strategy and Executive Order Seek to Deter Adversaries and Strengthen Resilience | Crowell & Moring LLP, accessed March 15, 2026, https://www.crowell.com/en/insights/client-alerts/white-houses-new-cyber-strategy-and-executive-order-seek-to-deter-adversaries-and-strengthen-resilience

21. New National Cyber Strategy and EO Lays Out a Path for Combating Cybercrime and Promoting Innovation - Wiley Rein, accessed March 15, 2026, https://www.wiley.law/alert-New-National-Cyber-Strategy-and-EO-Lays-Out-a-Path-for-Combating-Cybercrime-and-Promoting-Innovation

22. FINRA 2026 AI Governance: Managing Agentic and Shadow AI Risks - Smarsh, accessed March 15, 2026, https://www.smarsh.com/blog/thought-leadership/finra-2026-oversight-priorities-ai-communications-fraud/

23. March 2026 Regulatory Update: A $68M Fair Lending Settlement and More - Ncontracts, accessed March 15, 2026, https://www.ncontracts.com/nsight-blog/march-2026-regulatory-update

24. Horizon Scan AI and Deepfakes - FATF, accessed March 15, 2026, https://www.fatf-gafi.org/en/publications/Methodsandtrends/horizon-scan-ai-deepfake.html

25. KYC Regulatory Trends for 2026: FATF, FCA & EBA Focus - smartKYC, accessed March 15, 2026, https://smartkyc.com/kyc-regulatory-trends-to-watch-in-2026/

26. FATF plenary February 2026: Key grey list changes, new strategic publications, and a new presidential appointment - ComplyAdvantage, accessed March 15, 2026, https://complyadvantage.com/insights/fatf-plenary-february-2026-key-updates/

27. Rethinking Watchlist Screening Solutions for 2026 | Unit21 - Blog, accessed March 15, 2026, https://www.unit21.ai/blog/from-noise-to-precision-watchlist-screening-solutions-for-2026

28. INSIGHT: Agentic AI and stablecoins – the five trends redefining AML in 2026, accessed March 15, 2026, https://www.amlintelligence.com/2026/01/insight-agentic-ai-and-stablecoins-the-five-trends-redefining-aml-in-2026/

Disclaimer:

Fintech Wrap Up aggregates publicly available information for informational purposes only. Portions of the content may be reproduced verbatim from the original source, and full credit is provided with a “Source: [Name]” attribution. All copyrights and trademarks remain the property of their respective owners. Fintech Wrap Up does not guarantee the accuracy, completeness, or reliability of the aggregated content; these are the responsibility of the original source providers. Links to the original sources may not always be included. For questions or concerns, please contact us at sam.boboev@fintechwrapup.com.

Read more

Video of the Week

Deep Dive of the Week

Mastercard Verifiable Intent vs Visa Trusted Agent Protocol

Agentic commerce breaks a core assumption of online payments, that a human is directly clicking “buy” on a trusted surface. Once software can browse, decide, and transact, merchants and networks lose the simplest security primitive: “the customer was here.”

That creates three concrete trust failures that show up as operational costs:

Merchants need a way to distinguish a legitimate, user-authorized agent from malicious automation and bot traffic, without rewriting their stack or blocking valuable sessions by accident.

Payment networks and issuers need a deterministic audit trail of what the user authorized, what the agent did, and what the merchant charged, so disputes and fraud decisions can anchor to evidence rather than inference.

Agents need a portable way to carry proof of authority across merchants, devices, and protocols, in both human-present and human-not-present execution modes.

Both proposals are “trust layers,” but they sit in different parts of the stack and they optimize for different verifiers.

This week’s reports

1️⃣Sanctions in crypto just escalated again

2️⃣Stablecoins have moved from pilot to production in cross-border payments

3️⃣How Banks Are Building With AI

4️⃣SWOT Assessment of Agentic Commerce for Retailers

5️⃣Understanding The History of Agentic Payments

6️⃣Crypto Crime Is Becoming Financial Infrastructure

7️⃣AI Is Already the Operating System of Payments

This week’s insights

1️⃣An incomplete map of the crypto card universe

2️⃣The Agentic Commerce Market Map by Artemis

3️⃣The neobank landscape is fragmenting and converging at the same time

4️⃣Understanding BNPL Landscape & Credit Dynamics

5️⃣Neobanks entering new markets will poach consumer deposits from banks

6️⃣Agentic Commerce and the Potential Collapse of Card Payment Economics

7️⃣Centralised decentralised finance (CeDeFi) platforms misprice risk at the ledger layer

An incomplete map of the crypto card universe

Look at the landscape and one thing becomes obvious: this isn’t a niche anymore. It’s an ecosystem.

You have infrastructure players like Rain and Bridge sitting alongside program managers like Baanx and Gnosis Pay. You have wallets, onramps, settlement layers, and regional issuers. Cardano-native initiatives appear next to multi-chain stacks. Kulipa and others focus on enabling issuance-as-a-service.

This is no longer “a crypto card.”

It’s:

– Stablecoin funding rails

– Program management and compliance

– BIN sponsorship

– Wallet orchestration

– FX and settlement

– Traditional card network connectivity

In other words, it’s a full-stack financial product wrapped in a card form factor.

Read more

Agentic commerce breaks a core assumption of online payments, that a human is directly clicking “buy” on a trusted surface. Once software can browse, decide, and transact, merchants and networks lose the simplest security primitive: “the customer was here.”

That creates three concrete trust failures that show up as operational costs:

Merchants need a way to distinguish a legitimate, user-authorized agent from malicious automation and bot traffic, without rewriting their stack or blocking valuable sessions by accident.

Payment networks and issuers need a deterministic audit trail of what the user authorized, what the agent did, and what the merchant charged, so disputes and fraud decisions can anchor to evidence rather than inference.

Agents need a portable way to carry proof of authority across merchants, devices, and protocols, in both human-present and human-not-present execution modes.

Both proposals are “trust layers,” but they sit in different parts of the stack and they optimize for different verifiers.

Mastercard Verifiable Intent as a credential chain for authorization proof

Verifiable Intent (VI) is built as a layered credential format: a cryptographic delegation chain that binds identity, intent, and action using SD-JWT and key binding, with role-scoped selective disclosure.

The spec frames two execution modes:

Immediate mode is human-present. The user signs final checkout and payment values. The specification treats this as a two-layer flow, with short-lived Layer 2 credentials.

Autonomous mode is delegated. The user signs constraints up front, binds an agent key, and the agent later produces fulfillment credentials within those bounds. The spec treats this as a three-layer flow and makes constraints normative in that path.

The mechanics matter because Verifiable Intent is not just an “agent header.” It is a composable evidence object that multiple parties can verify independently.

• Layer 1 is an SD-JWT issued by an issuer and provisioned into a credential provider wallet, binding user identity to a public key via cnf.jwk. The normative format specifies ES256 and a long lifetime on the order of a year, plus issuer key discovery through JWKS.

• Layer 2 is created and signed by the user key bound in Layer 1. It expresses purchase intent either as final values (Immediate) or as constraint-bearing mandates and an agent key binding (Autonomous). Layer 2 also binds to Layer 1 via sd_hash.

• Layer 3 exists only in Autonomous mode. The agent signs short-lived, key-bound SD-JWTs that present fulfillment details. The format splits L3 into a network-facing payment mandate (L3a) and a merchant-facing checkout mandate (L3b). The split is not aesthetic; it enforces a privacy boundary by construction.

Constraints are treated as first-class authorization limits. The constraints spec defines eight registered constraint types and requires verifiers to support them. It also defines how unknown constraint types behave under strictness modes, and it explicitly calls out that unknown types in open mandates must be rejected because they can silently unbound authority.

This is where Verifiable Intent is opinionated: it tries to turn “intent” into machine-checkable bounds, and it makes the verifier do explicit constraint validation against fulfillment artifacts like the merchant-signed checkout object.

Read more