Agentic Bank Payments Reach the UK Under FCA Supervision

GoCardless has processed what it describes as the UK's first live agentic bank payment, a recurring charity donation set up from start to finish by an AI agent operating under Financial Conduct Authority supervision. The payment, a monthly gift to the anti-poverty charity Trussell, was completed inside a conversational checkout: an AI agent answered questions about the charity's work, presented monthly giving options of £5, £10 or £15, then set up a Direct Debit mandate once the donor chose an amount. It is an early production test of account-to-account rails as the infrastructure for agentic commerce, the emerging model in which AI agents buy and pay on a customer's behalf within limits the customer sets.
The demonstration is small in monetary terms, but the timing is deliberate. GoCardless, now a Mollie company following the completion of a roughly €1.05bn acquisition on 1 September 2026, is trying to establish bank payments rather than cards as the default rail for AI-driven purchases before the category scales.
What actually happened in the transaction?
The entire interaction took place within an AI chat rather than a conventional web form. A donor could ask about Trussell, review how each donation tier would be used, and confirm a monthly commitment, at which point the agent prompted for bank details and created the Direct Debit mandate. GoCardless says the flow was designed so the payer keeps control at every step, reflecting its own consumer research in which 64% of UK adults surveyed said they were open to AI managing recurring payments provided they retained control over limits or final approval. That figure is a company-stated survey result rather than an independent finding.
Why is GoCardless backing Direct Debit for agentic payments?
The strategic argument is about fit. Direct Debit and recurring open banking payments are built around standing consent and repeat collection, which maps onto the kind of recurring, rules-based spending that consumers appear most willing to delegate to an agent. Cards, by contrast, were designed for one-off authorisation. GoCardless has argued for some time that account-to-account payments are the natural foundation for agentic commerce, and this transaction is intended to convert that positioning into a working reference case.
That framing sits against a specific commercial backdrop. GoCardless reported £126.8m in revenue for its 2024 financial year, up 38% year on year, and recorded its first EBITDA-positive quarter in summer 2025 before the Mollie deal closed. An early lead in agentic account-to-account payments would give the combined group, which serves more than 350,000 businesses across 30-plus markets, a point of difference against card-centric competitors.
How big could agentic commerce actually become?
The prize is significant but still speculative. Morgan Stanley Research estimated in December 2025 that agentic shoppers could account for between $190bn and $385bn of US e-commerce spending by 2030, equivalent to 10% to 20% of the online retail market. The $385bn figure GoCardless cites is the top, bull-case end of that range rather than a central estimate, a distinction worth holding onto given how early adoption remains. Most of the agentic-commerce infrastructure announced by large networks and processors to date has been built around cards, which makes an account-to-account entrant a notable counterpoint rather than the consensus approach.
Why does FCA supervision matter here?
GoCardless built the checkout inside the FCA's AI Live Testing programme, a supervised scheme launched in 2025 that lets firms trial consumer-facing AI in live markets with regulatory and technical oversight. The regulator's first named cohort in December 2025 included NatWest, Monzo, Santander and Scottish Widows, with a second intake beginning testing from April 2026. Running an agentic payment under supervision gives the exercise a governance story that a standalone product launch would lack, and that matters because agent-initiated payments raise unresolved questions about consent, liability and consumer protection. Card purchases carry Section 75 protection in the UK and Direct Debit carries its own guarantee, but the industry has yet to settle who is accountable when an agent acts beyond a customer's intent.
The policy direction is supportive. HM Treasury's National Payments Vision, published in November 2024, set an explicit ambition for seamless account-to-account payments to become a genuine alternative to cards, with open banking central to delivering it. Agentic use cases hand that ambition a concrete, and politically useful, proving ground.
Why This Matters to FinanceX Readers
For payments professionals, the signal is that the contest over agentic commerce rails is now live rather than theoretical, and that account-to-account providers intend to compete for it rather than cede the ground to cards. A single charity donation will not shift volumes, but it stakes a claim: if consumers delegate recurring payments to agents first, as GoCardless's research suggests, the rails built for standing consent could capture that behaviour early.
For investors, the watch items are concrete. Track whether agentic account-to-account moves beyond supervised pilots into commercial volume, how quickly the card networks respond with their own bank-payment agent flows, and whether regulators clarify liability for agent-initiated transactions. Each will shape which payment rails, and which providers, command pricing power as agentic spending grows from demonstration to habit.



