One Year, 40,000+ Subscribers: Agentic Payments, Kill Switches, and the AI Org Chart Rewrite...
Welcome to Agentic Newsletter for Financial Services, the first newsletter dedicated to the transformation in financial services driven by the advancements in AI, shaping the Agentic economy and crafted by a team of more than half a dozen agents that keep evolving and adapting.
These Agents scan the world on a daily basis, filtering hundreds of news articles. We work together to bring you a weekly curated selection coupled with my human insights.
Thanks for reading and being part of our fast-growing subscription community.
This is edition No. 56. Celebrating one full year and crossing the 40,000 subscribers milestone recently.
Let's stay up to date and build a more human world powered by agentic AI. Thank you for reading.
The plumbing for Agentic Payments competition
Cross River Bank and Stripe formalized bank-grade card issuance that lets agents spend through money that already exists, just as Robinhood built a purpose-made credit line for agentic spend, and BNB Chain gave agents independent on-chain wallets and IDs outside banking plumbing entirely.
On the governance side, the FCA's Mills Review proposes extending the Critical Third Parties regime to reach AI model developers and cloud providers directly, not just the banks that deploy them, while the Bank of England and India's RBI joined the OCC and Fed in making kill switch capability a supervisory expectation. Three regulators, three jurisdictions, the same mechanism, arriving within weeks of each other.
Meanwhile, S&P Global restructured its entire Market Intelligence division in what looks like an AI native reorganization versus a product based organizational chart.
AGENTIC COMMERCE, RAILS & INFRASTRUCTURE
Two designs for how AI agents pay
What Happened: Cross River Bank and Stripe formalized bank-grade card issuance for AI agents — infrastructure that works across Visa and Mastercard rather than a single network. The design uses virtual, single-use cards that draw on whatever payment method - debit, credit, or linked bank account, a user already has in their Stripe wallet, so software can spend without ever touching the underlying credentials.
Robinhood's Agentic Credit Card, launched with Visa and Coastal Community Bank a month ago is different. Robinhood built a new credit product for agents to spend against, while Cross River and Stripe built a pass-through layer for agents to spend through money that already exists. The latter is infrastructure.
Days later, BNB Chain and AWS launched BNB Agent Studio on Bedrock AgentCore, letting agents hold their own onchain wallets under ERC-8004 identity and reputation registries.
Why All This Matters: Same problem, three architectures. Cross River keeps the agent inside the existing card network and its regulatory perimeter — dispute resolution, liability, fraud rules all inherited from decades of card law. Robinhood extends that same perimeter but with a purpose-built credit line instead of a wrapper on existing funds. BNB Chain gives the agent an independent, self-sovereign identity outside banking entirely.
My guess is all three coexist for the next five years, split by use case: the delegate economy (a consumer's agent doing errands and shopping) runs on Stripe/BaaS-style infrastructure, while the machine-to-machine economy (high-volume microtransactions between agents and services) runs on-chain, where settlement finality and programmability matter more than dispute resolution.
Visa quietly standardized agentic payments across a dozen European banks at once
What Happened: At the Visa Payments Forum in Paris on July 2, Visa revealed that its Agentic Ready programme now spans Barclays, BBVA, CaixaBank, Commerzbank, Bank of Cyprus, HSBC UK, ING, Klarna, Lloyds, mBank, Nationwide, NatWest, Nexi Group and Revolut — all running live AI agent-initiated transactions on the same infrastructure: Visa Intelligent Commerce for tokenisation and fraud monitoring, and Visa Payment Passkeys for EU Strong Customer Authentication compliance. BBVA and CaixaBank were simply the two who went public with named case studies first.
Why It Matters: This isn't twelve banks independently proving agentic payments work. It's Visa running one standardized rollout and letting the participants take turns claiming "first." That's a materially different signal than the piecemeal "Bank X completes its first agentic transaction" headlines suggest. Worth watching which of the twelve banks moves into actual transaction volume first.
MARKET IMPACT, FUNDING & INNOVATION
S&P Global rebuilds itself around Kensho
What Happened: S&P Global reorganized its Market Intelligence division into Kensho Data & Platforms and Enterprise Solutions, consolidating Capital IQ, Ratings Direct, Visible Alpha and With Intelligence under one AI-native structure with named leadership reassignments across the business.
Why It Matters: S&P isn't defending a data moat by sitting still. It's restructuring the org chart so Kensho becomes the delivery layer for everything else. Four previously separate brands/product lines folding into two AI-native divisions, with named leadership reassignments (Sally Moore, Darren Thomas, Bhavesh Dayalji, Whit McGraw) sitting underneath.
REGULATION & GOVERNANCE
The FCA's Mills Review roadmap
What Happened: Drawing on nearly 140 industry submissions and a 5,000+ consumer survey, the FCA's Mills Review concludes AI will reshape UK retail financial services by 2030, with roughly 11 million adults likely to use autonomous agents.
Rather than write AI-specific rules now, it lays out a seven-point roadmap: scaling the FCA's AI Lab, adapting the regulatory perimeter, strengthening system-wide oversight, monitoring autonomous models, building agentic finance foundations, a public-interest AI capability service and, most consequentially, expanding the Critical Third Parties regime to bring major AI model developers and cloud providers under direct regulatory reach, not just the banks that use them.
Why It Matters: That last point is the one worth sitting with. Right now the FCA can examine how a bank uses AI, but it can't touch OpenAI, Anthropic or a hyperscaler directly — the CTP expansion closes that gap.
#AndTheIronyIs the review that declined to write new AI rules just proposed the regulatory mechanism most likely to reshape who's actually accountable when the model, not the bank, is the point of failure.
The kill switch is becoming an industry standard, one regulator at a time
What Happened: After the OCC and Fed made kill-switch capability mandatory in bank examinations two weeks ago, the Bank of England's Sarah Breeden and India's RBI are following with their own versions — Breeden warning that human oversight of every agent action "is unlikely to be realistic," and the RBI's draft requiring board-level accountability and documented kill-switch capability across banks, NBFCs and credit bureaus.
Why It Matters: Three regulators, three jurisdictions, the same mechanism. When US, UK and Indian supervisors converge on identical language within a month of each other, that's not coincidence. It's a sign the industry itself is quietly lobbying for a shared standard because nobody wants to be first to explain why they can't switch an agent off. However, the US is asking about it, while the UK and India are moving to require it.
The kill-switch convergence is the detail worth dwelling on. A kill switch presumes you can see what you're stopping. When agents interact with other agents across institutions, the failure mode isn't one model misbehaving — it's coordination no single firm can observe from its own logs. The supervisory question is what evidence of inter-agent behaviour a regulator could actually inspect after the fact.
Congratulations on the one year milestone and 40,000 subscribers! 🎉 The regulatory convergence around kill switch requirements is the most critical piece here. Three major supervisors aligning on the same supervisory requirement within weeks shows how quickly governance frameworks are adapting to agentic payments. From a governance and compliance perspective, bank and card rails offer the clearest path for accountability. They operate within established regulatory perimeters, making liability and audit trails much easier to enforce. Onchain wallets introduce incredible autonomy but present a massive challenge for compliance and risk management. I would lean toward bank and card rails for the built-in oversight and control mechanisms. #DeliverRightResults #GovernanceMatters #AgenticPayments
Many congrats Dr. Efi Pylarinou, great insights so it is well deserved, onwards and upwards!
An interesting pattern is emerging across these developments. When independent regulators begin converging on similar supervisory mechanisms, it often reflects a deeper architectural convergence rather than simple regulatory alignment. As autonomous systems become part of critical infrastructures, governance appears to be evolving toward shared architectural primitives that can operate across jurisdictions, technologies and implementation models. The question may gradually shift from which governance framework is adopted to which governance structures remain interoperable across increasingly heterogeneous ecosystems.
The technology is moving quickly, but the real differentiator will be governance. As AI agents begin transacting autonomously, trust will depend on clear authority, accountability, and intervention mechanisms.