AI For Payments Firms: UK Guide To Agentic Commerce In Plain English

What is agentic commerce?

Agentic commerce is shopping and paying done by an AI agent on behalf of a person or a business. The person sets the goal and the limits, such as “renew this subscription if the price stays under £20”. The agent finds options, compares them and completes the purchase within those limits.

NatWest describes it as “a future where AI agents could play a greater role in how people choose and pay for products and services” (NatWest Group, 22 September 2026). HM Treasury's July 2026 consultation describes the payment side as “The use of agentic AI to autonomously analyse, initiate, approve, and execute payments on behalf of consumers or firms”.

How does agentic commerce work?

The details vary by provider, but most designs follow the same steps:

1. The customer gives the agent a goal and limits, such as a budget, a list of approved merchants or a date.
2. The agent searches, compares and picks an option.
3. The agent starts a payment using a payment method the customer has authorised for it.
4. The payment provider authenticates the transaction and checks it against the customer's consent.
5. The merchant and the payment firms process, settle and reconcile the payment, as with any other transaction.

The hard questions sit in steps three and four: what the customer actually consented to, how the agent proves who it is, and who pays if it gets something wrong.

What did NatWest and five other banks agree?

On 22 September 2026 NatWest Group, ASB Bank, Bank of America, Capital One, Commonwealth Bank of Australia and ING Group published principles for trusted agentic commerce. NatWest said the principles “focus on five areas: transparency, safety, privacy and data, choice and interoperability.”

The aim, NatWest said, is that customers and merchants “retain choice, control and flexibility in how they pay and get paid.” The banks presented the paper as “a foundation for further discussion” and invited others to help apply it in practice.

Customers need to trust that they remain in control of how payments are made and that their money is safe.

- Mark Brant, Chief Payments Officer, NatWest Group, 22 September 2026

Is agentic commerce regulated in the UK?

Agentic payments run on the existing payment rules, and those rules were not written with agents in mind. HM Treasury's consultation on modernising payment services regulation, published in July 2026, says: “the PSRs were designed before the development of AI and may not fully facilitate the use of agentic AI.”

Its Question 15 asks: “How does existing payment services regulation need to adapt to support agentic payments? For example, do provisions relating to authentication and consent of payments transactions, and liability for unauthorised payment transactions, need updating?” The consultation was open for 12 weeks and closed on 6 October 2026. The Treasury has not yet published its response.

The Treasury's Financial Services AI Adoption Plan, published the same day in July, recommends using that consultation to build a trust framework for agentic payments. One of its three pillars is “Know Your Agent” protocols: “standardised identity and verification frameworks specifically designed for AI and autonomous software agents.”

The FCA has said it is looking at the same question. Its Payments Regulatory Priorities report of March 2026 says its work to modernise payment regulation “will include considering whether change or development of regulation is needed to support agentic AI payments.”

Who is liable if an AI agent makes the wrong payment?

That question is open. Liability for unauthorised transactions is one of the specific points the Treasury consultation asks about, and the answer will depend on the government's response and any later rules. Firms should take legal advice on their own position. This article is not legal advice.

What does agentic commerce mean for UK payments and e-money firms?

Long before the rules settle, agents will change the shape of the work in operations. More payments may be triggered by software, at odd hours and in patterns that differ from human behaviour. Each one still has to be authorised, recorded, settled and reconciled.

• Consent records. Firms will need to show what a customer authorised an agent to do, and when.
• Exceptions. Declines, disputes and mismatches from agent-initiated payments will land in the same queues as everything else.
• Reconciliations. Higher volumes put more weight on matching every movement, every day.
• Safeguarding. Since 7 May 2026 payment and e-money firms have had to carry out daily checks that the right amount of customer money is safeguarded (FCA).

Where does AI for payments firms in the UK pay back first?

Safeguarding is already a strain for many firms. Kani Payments' safeguarding readiness research, published on 6 May 2026 and based on 75 compliance and finance professionals at FCA-regulated payment and e-money institutions, found that just 13% were performing daily safeguarding reconciliations, and that 64% still relied on spreadsheets for monthly safeguarding returns.

This is where AI pays back first. An agent can prepare the daily safeguarding reconciliation and write up every difference, and a person approves it. Our agents never move money. See how we approach safeguarding reconciliation automation, and how reconciliation agents are changing exception handling.

Where should a payments firm start?

Get the basics of today's volumes right before agent-initiated payments add to them. BraivIQ, an AI agency in the UK for financial firms, starts with a 14-day Proof Run on the firm's own exports, read-only, which finds every mismatch and times the manual work on real cases.

For teams following agentic AI in London and the wider UK market, the questions to settle early are the ones the banks and the Treasury are asking: what the customer consented to, how an agent is identified, and how every movement is reconciled.