What Are Agentic Payments? The Rise of AI That Shops, Compares, and Pays Without You

What Are Agentic Payments? The Rise of AI That Shops, Compares, and Pays Without You

Imagine waking up on a Monday morning to find that your business received three new orders overnight. No human placed them. No one browsed your product pages, weighed up the options, or typed in a card number. An AI agent, working on behalf of a customer who set it loose the previous week with a budget, a set of preferences, and a list of approved suppliers, found you, compared your prices against two competitors, decided you offered the best value, and completed the purchase. The confirmation email is already in your inbox.

This is not a thought experiment. It is the direction the global payments industry is moving in, and it is moving fast.

Agentic payments are the mechanism by which artificial intelligence systems make financial transactions autonomously, without a human approving each individual step. The concept sits at the intersection of AI, digital identity, and payments infrastructure. It is also one of the most consequential shifts in commerce since the introduction of contactless cards. And right now, most UK small business owners have never heard of it.

That needs to change.

---

First, What Is an AI Agent?

Before we get to the payments part, it helps to understand what an AI agent actually is.

A standard AI tool, think ChatGPT in its basic form, responds when you ask it something. You prompt it, it answers, and it stops. An AI agent is different. It is a system designed to pursue a goal across multiple steps, taking actions in the real world to achieve that goal, without needing you to hold its hand at every stage.

You might tell an AI agent: "Book me the cheapest return flight to Edinburgh in the first two weeks of March, make sure it is not before 8am, pay with my Amex, and send the confirmation to my work email." The agent then goes and does exactly that. It searches, compares, selects, and transacts. You are not there for the payment step. You authorised the outcome at the start, not each individual click.

Now scale that logic up. Businesses are beginning to use AI agents to manage procurement, reorder stock, book logistics, pay supplier invoices, and compare service contracts. Consumers will use them to manage subscriptions, book appointments, order groceries, and compare insurance quotes. The agent acts. The payment follows automatically.

That payment step is what the industry is now scrambling to build proper infrastructure around.

---

Why Payments Infrastructure Had to Catch Up

Here is the problem that emerged as AI agents became more capable. The payments system was built for humans. Card networks, authentication protocols, fraud detection, even the basic idea of cardholder consent: all of it was designed on the assumption that a person was present, or at least closely involved, in every transaction.

Agentic payments break that assumption entirely.

If an AI agent places an order on your behalf, who is the cardholder? Who consents to the charge? If the agent makes an error, or gets manipulated by a malicious third party into redirecting a payment, who is liable? And how does a merchant's terminal or payment gateway even verify that the entity completing a transaction is a legitimate, authorised AI system rather than a fraudulent bot?

These are not abstract philosophical questions. They are technical and regulatory problems that the card networks have had to confront urgently, because the agents are already here. The infrastructure to support them safely is still being built.

---

Visa's Answer: The Intelligent Commerce Initiative and TAP

In April 2025, Visa announced what it called its Intelligent Commerce initiative, a comprehensive programme to rebuild the payments rails for an AI-first world. The centrepiece is a protocol called TAP, which stands for Token Agent Protocol.

TAP is Visa's framework for issuing payment credentials specifically to AI agents rather than to human cardholders. Here is how it works in simplified terms.

A consumer or business authorises an AI agent by setting parameters: a spending limit, approved merchant categories, a time window, perhaps a list of blocked vendors. Visa then issues a tokenised credential, a unique payment identifier, to that agent. The token carries the spending rules baked into it. When the agent goes to make a purchase, it presents the token, the merchant's system verifies it through Visa's network, and the transaction is approved or declined based on the pre-set parameters.

Critically, the merchant never sees the underlying card details. The agent never has unrestricted access to the account. And every transaction is traceable back to the original authorisation event, which matters enormously for dispute resolution and fraud liability.

Visa has stated that it is working with over 60 technology partners on this initiative, including Anthropic, Microsoft, OpenAI, Stripe, and IBM. That is not a pilot list. That is the architecture of the next phase of commerce being assembled in real time.

---

Mastercard's Approach: Agent Pay

Mastercard has been equally direct. In early 2025, the company introduced Agent Pay, its own framework for enabling AI agents to transact securely within the Mastercard network.

Mastercard's approach leans heavily on its existing tokenisation infrastructure and on what the company calls "agentic commerce guardrails": a set of verification and consent mechanisms designed to ensure that every AI-initiated payment can be traced to a genuine human authorisation. Mastercard has framed this around three principles: that the human is always in control, that every transaction is transparent and auditable, and that the agent's credentials can be revoked instantly if something goes wrong.

Agent Pay is being integrated with Mastercard's broader identity and open banking capabilities, and the company is working with major AI platform developers to embed these protocols directly into agent frameworks at the development stage, rather than bolting them on afterwards.

What is notable about both Visa and Mastercard's approaches is that they are not treating agentic payments as a distant future problem. They are treating it as an immediate infrastructure requirement, because the agents are already transacting. The question is whether those transactions happen safely, with proper authorisation trails, or in a grey zone that creates liability and fraud exposure for everyone in the chain, including merchants.

---

What Does This Mean for UK Businesses Right Now?

Let us be honest about the timeline. If you run a coffee shop in Leeds or a photography supplies shop in Bristol, AI agents are not placing orders with you this week. The consumer-facing rollout of agentic commerce is still in early stages, and in the UK, the regulatory framework is still forming.

The Financial Conduct Authority and the Payment Systems Regulator have both signalled awareness of AI's growing role in financial services. The PSR's ongoing work on payment fraud liability and the FCA's AI and Digital Regulation discussions will both need to address agentic transactions explicitly. That work is underway but not yet concluded.

However, "not this week" does not mean "not soon". And for UK SMEs, the relevant question is not whether agentic payments will arrive but whether you will be ready when they do.

Here is what you should be thinking about now.

Tokenisation compatibility. If your payment system does not support tokenised transactions, you may find yourself unable to accept agent-initiated payments when they become mainstream. Ask your payment provider where they stand on Visa TAP and Mastercard Agent Pay compatibility. A provider that cannot give you a clear answer is a provider that is behind.

Authorisation and dispute processes. When a transaction is initiated by an AI agent and something goes wrong, the dispute resolution process looks different. You will want to know that your payment provider has a clear position on liability in agent-initiated transactions, and that your own records are clean enough to defend a chargeback.

Fraud pattern changes. AI agents are also available to bad actors. Fraudulent agents, systems designed to probe for weaknesses in merchant checkout flows or exploit authorisation gaps, will become a real threat. The fraud patterns you are used to defending against will evolve. Staying close to your payment provider's fraud detection updates will matter more, not less.

B2B procurement is the early frontier. The first wave of agentic payments in the UK is most likely to come through business-to-business channels. If you sell to other businesses, particularly if you operate in sectors like office supplies, logistics, software, or professional services, an AI procurement agent may be evaluating your pricing and terms right now. Your online presence, pricing transparency, and API accessibility will start to matter for the AI audience, not just the human one.

---

The Bigger Picture: Commerce Is Changing at the Decision Layer

The most profound implication of agentic payments is not really about payments at all. It is about where purchasing decisions get made.

For decades, businesses competed for human attention. Your shop window, your Google ranking, your product photography, your customer reviews: all of it was designed to influence a human decision-maker at the moment of consideration. Agentic commerce shifts that decision-making to a layer that operates before the human is even involved.

An AI agent shopping on behalf of a consumer will be optimising against a set of parameters: price, delivery speed, return policy, merchant reliability score, perhaps carbon footprint. It will not be moved by your shop window or your brand story in the same way a human browser might be. It will read your structured data, your pricing API, your product specifications, and your historical fulfilment record.

This does not mean brand and experience become worthless. It means the game expands. You still need to win the human. But increasingly, you will also need to be legible to the machine.

---

Practical Takeaways for Your Business

Ask your payment provider directly: Do your payment terminals and online gateway support tokenised, agent-initiated transactions? What is their roadmap for Visa TAP and Mastercard Agent Pay?

Audit your product data: If an AI agent is going to evaluate your business, your pricing, stock levels, product descriptions, and delivery terms need to be accurate, structured, and machine-readable. A messy website with inconsistent information will lose agent-initiated customers before you ever knew they existed.

Review your fraud thresholds: Talk to your payment provider about whether your current fraud settings are calibrated for automated transaction patterns. Agent-initiated purchases may look different from human-initiated ones in terms of timing, frequency, and basket composition.

Stay close to FCA and PSR updates: The regulatory framework for agentic payments in the UK is still forming. Follow PSR consultation publications and FCA AI-related guidance. These will shape your liability exposure and your compliance obligations.

Do not wait for certainty: The businesses that will handle agentic commerce well are those that begin understanding it now, before it is mainstream, not those that scramble to adapt once it has already changed their market.

---

The payments industry has been here before. Contactless seemed futuristic until it was everywhere. Open banking seemed abstract until it powered the apps your customers use daily. Agentic payments are at that same inflection point, somewhere between emerging and inevitable.

The AI is already shopping. The infrastructure to support it is being built right now by the largest card networks on the planet. The only question is whether your business is positioned to receive it.

Sources

1. Visa Intelligent Commerce initiative announcement, April 2025: https://investor.visa.com/news/news-details/2025/Visa-Introduces-Visa-Intelligent-Commerce/default.aspx
2. Mastercard Agent Pay announcement and agentic commerce framework, 2025: https://www.mastercard.com/news/press/2025/february/mastercard-launches-agent-pay-pioneering-the-future-of-agentic-commerce/
3. BIS CPMI working paper on AI in payments infrastructure and settlement systems: https://www.bis.org/cpmi/index.htm
4. Payment Systems Regulator (PSR) strategy and AI-related payment fraud consultation documents: https://www.psr.org.uk/publications/
5. Financial Conduct Authority (FCA) AI and Digital Regulation Service guidance: https://www.fca.org.uk/innovation/ai-digital-regulation-service
6. McKinsey Global Payments Report 2024, section on embedded finance and AI-driven commerce: https://www.mckinsey.com/industries/financial-services/our-insights/the-2024-mckinsey-global-payments-report
7. PYMNTS.com coverage of agentic commerce and AI-initiated transaction volumes: https://www.pymnts.com/artificial-intelligence/2025/
8. Fintech Brain Food by Simon Taylor, newsletter coverage of agentic payments and UK fintech implications: https://www.fintechbrainfood.com
9. The Paypers coverage of tokenisation standards and AI agent payment rails: https://thepaypers.com
10. Anthropic partnership announcements related to Visa Intelligent Commerce programme, April 2025

---

Disclaimer

The views and information shared in this post are for educational and informational purposes only and do not constitute financial, legal, or professional advice. While every effort is made to ensure accuracy, Klipy UK Limited accepts no liability for decisions made based on this content. Payment processing rates, regulations, and product features referenced are subject to change. Klipy UK is an authorised seller of Teya payment solutions. Where third-party sources are cited, links are provided for reference; Klipy UK does not endorse or guarantee the accuracy of external content. For personalised guidance on your business payment needs, please contact us directly at editor@klipy.uk.