Digital payments have long assumed that a person chooses a purchase and initiates or approves the transaction. But today, agentic commerce challenges that foundation by allowing software to act within delegated authority.

That shift could reshape not only checkout, but also the allocation of trust, control and value across the payments ecosystem.

Increasingly capable AI agents can search for products and services, compare options, negotiate terms and execute transactions within customer-defined parameters.

Could the next evolution of digital commerce be less about simplifying individual payments and more about removing the need for people to initiate them one by one?

Let’s explore the possible outcomes and implications of agentic commerce.

In simple terms, agentic commerce is software that acts on a customer's behalf to complete commercial tasks within a defined mandate.

Unlike an AI assistant, AI agents have autonomy to take certain actions (they have agency, hence ‘agent’). For example, and AI Assistant might be able to find you the cheapest flights to New York, but an AI agent could have the authority and ability to not only find the flight, but also book based on your predetermined preferences, and complete the payment within an agreed spending limit.

It’s a subtle but important distinction that could have significant implications for financial services and fintechs. Why? Because the moment an AI system can execute a financial transaction, questions of authorisation, authentication, liability, fraud and regulation become central.

For example, who would be liable if an agent makes a purchase in error? And how might fraudulent agents be detected and stopped? And what’s the role of people in all of this?

These aren’t hypothetical questions. In September 2026, Reuters reported that India was preparing a framework under which AI agents could make certain low-value payments through UPI without approval for every transaction¹. The proposed controls included spending limits, identity checks, rule-based instructions, audit trails and a liability framework.

That’s a significant step. UPI is the world's largest retail fast-payment system by transaction volume, so adding agentic functionality to national infrastructure would provide a powerful real-world test of how autonomous software operates on mainstream payment rails.

It’s easy to imagine how the shift to agentic commerce could play out.

Rather than opening a retailer's website, searching through products, comparing prices, entering payment information and completing checkout, a customer could simply delegate the task to an agent.

That agent can then continuously monitor prices and make purchases according to predefined rules – plus it can keep a record of what it’s done, and why.

While that sounds great for customers because it means they don’t have to do the “heavy lifting” of buying online, it’s a change that has more serious implications for merchants and payment providers – especially if they’re already deploying their own agents.

In an agentic model, interactions could increasingly occur between an AI agent and the merchant's systems, with the human specifying the objective rather than personally executing every step.

With agents making decisions based on pre-determined criteria, customer relationships could be formed in entirely new ways, with value accruing in different parts of the payment ecosystem.

Would a business even need persuasive marketing if a machine is automatically making decisions? And how might a business ‘convince’ an agent that their offering is superior to a competitor?

There are certainly questions that need to be addressed – including questioning the strength of regulation around rapidly-advancing tech.

The primary regulatory challenge around agentic commerce is that existing payment frameworks were largely designed around human decisions and clearly identifiable payment instructions.

When an agent is acting on behalf of a person or business, the financial institution needs to establish not only who ultimately owns the account, but what authority has been delegated to the agent and whether the specific transaction falls within that authority.

Imagine a corporate treasury team giving an AI agent authority to purchase foreign currency up to a specified limit. If the agent makes a transaction outside those parameters, who is responsible? What if the agent has been manipulated? What if it interprets an instruction incorrectly? What evidence would a bank need to demonstrate that the transaction was properly authorised?

These questions are not peripheral governance concerns; they determine whether agentic payments can be authorised, supervised and trusted at scale.

The FCA has identified agentic systems as a material development for financial services², including systems that coordinate and transact. Its current approach remains principles-based and outcomes-focused, but it has emphasised that accountability for regulated activities must remain clear and that human oversight, consumer protection and resilience cannot be weakened.

This means that the critical infrastructure for agentic payments will not simply be the technology that allows an AI to press the equivalent of a "Pay" button. It will be the infrastructure that establishes what the agent is permitted to do.

Identity will matter. Authentication will matter. Spending limits will matter. Audit trails will matter. Real-time fraud monitoring will matter. And perhaps most importantly, there will need to be a clear, enforceable link between the actions of an autonomous system and the human or organisation that ultimately delegated authority to it.

This is where the fintech opportunity becomes particularly interesting.

Financial institutions have spent years building infrastructure around customer authentication and transaction monitoring. Agentic payments potentially require a new layer on top of that infrastructure: systems capable of establishing an agent's identity, verifying its mandate, evaluating the transaction against predefined rules and recording the basis on which the transaction was permitted.

The implications extend well beyond consumer shopping. Corporate procurement, travel management, expense administration, treasury operations, insurance and subscription management could all eventually involve software making transactions according to predefined objectives.

Agentic commerce represents a fundamental shift from payments initiated by humans to payments orchestrated by software on behalf of humans.

This is happening now, and it can have a profound impact on fintechs, who should treat agentic payments as an infrastructure and governance priority now.

Although the technology remains immature, and security, liability, consumer protection and regulatory questions are unresolved; greater AI capability alone does not guarantee reliable, controlled action.

The strategic question is no longer simply how AI can improve the customer experience, but how financial-services architecture must evolve when customers delegate decisions and transactions to software.

The future of payments may not be defined by making checkout faster.

It may be defined by making checkout unnecessary.

And if that happens, the companies that succeed may be those that build the trust, control and financial infrastructure that allow machines to act on our behalf without removing human accountability.