AI agents at the checkout: Why the UK payments sector must prepare for agentic commerce

Welcome to the age of agentic autonomous commerce. The future isn’t years away, but arriving at AI speed means months away.

In early 2024, Mastercard quietly unveiled “Agent Pay” and its vision for autonomous AI-driven transaction processing. Visa quickly followed with its “Intelligent Commerce” Initiative. These aren’t isolated moves but signal a major paradigm shift across the industry.

AI agents are no longer confined to recommendation engines or scripted chatbots. They are becoming autonomous actors, capable of making financial decisions and executing payments without human initiation and on behalf of individuals.

This marks the dawn of agentic commerce, a shift poised to become the most significant disruption to the payments sector in decades.

The AI commerce revolution

The race toward dominance in agentic commerce is on! This race will profoundly change how consumers shop, businesses manage procurement cycles, and commerce and payments function at their core.

Momentum is rapidly building among key players. AI vendors, such as Openai, Google, Anthropic’s Cloud, Mistral, and Meta, are advancing agentic AI technologies rapidly. Major retailers are starting to move to ensure they remain relevant and can capitalise on new ways to engage consumers and corporations.

Meanwhile, the payments industry must urgently assess the disruption ahead: how to adapt, where to intervene, and how to turn this massive transformation into an opportunity instead of a threat, as the current payment infrastructure is not built for AI actors.

The threat and opportunity are huge, at the intersection of two powerful markets: a projected $50 billion AI agent sector and the $36.75 trillion digital payments industry, redefining everything from B2C shopping to B2B procurement.

The Infrastructure Challenge

Our current payment infrastructure, designed for human-initiated transactions, is fundamentally unprepared for autonomous AI actors. Existing payment rails lack native support for critical new requirements: secure agent identity, behaviour-based authorisation, and AI-specific fraud detection.

We must prioritise the end-users. Will individuals and businesses comfortably delegate purchasing power to AI agents without robust, unified control and visibility? Likely not. What’s needed is an industry-wide approach to keep users in the loop, ensuring unified control, visibility, and oversight over AI agents’ activities across the ecosystem.

Furthermore, fraud detection frameworks are designed for existing payment rails and are inadequate for Autonomous commerce transactions. AI agents will introduce new transactional, contextual and behavioural patterns.

Standardised methods to authenticate, authorise, and verify these autonomous agents are essential to prevent tampering and impersonation and are crucial, as new risks will emerge at a fast pace as autonomous commerce expands.

Critically, we must acknowledge the historical inertia of payment infrastructure. As experiences with PSD2/3 demonstrated, core changes are slow, expensive, and resource-intensive. The payment industry needs to find a way to adapt quickly and become an enabler of autonomous payments. Otherwise, we may witness a wave of new players entering the space with alternative solutions, leading to fragmentation and chaos and displacing existing incumbents and regulators likely to play catch-up, responding to market innovation with reactive policy.

Orchestration: A secure path forward

One approach to transforming the current payment rails to enable Autonomous payments could be establishing a standard agentic AI orchestration layer that would enable existing payment gateways and infrastructure to support autonomous transactions securely and efficiently without requiring a complete overhaul. This approach can align with current regulatory frameworks while extending system capabilities for AI-driven commerce.

A potential and promising approach to enabling autonomous payments is creating a standardised agentic AI orchestration layer. This layer would allow existing payment gateways and infrastructure to support AI-driven transactions securely and efficiently while minimising changes and the impact on current gateway systems.

This orchestration layer extends the capabilities of current systems and aligns with regulatory frameworks, offering the industry a practical and scalable path forward.

A robust orchestration framework for autonomous payments should deliver the following:

• Verifiable agent identity: Strong authentication and authorisation mechanisms to prevent AI agent impersonation and ensure origin integrity.
• User-controlled agentic wallets: Standardised systems allowing users to oversee and manage all AI agents acting on their behalf, with enforceable contextual and behavioural rule sets.
• Contextual and behavioural authorisation: New authorisation flows are designed explicitly for AI agents, enabling transactions to be approved or rejected based on dynamic, context-aware, and behavioural rules.
• Autonomous agent fraud detection: Systems capable of analysing agent behaviour patterns in real-time and historically to detect fraud, impersonation, and anomalies.

A call for action for the payment industry

To successfully transition into the era of agentic payments, the global payments ecosystem must act collaboratively and at a fast pace:

1. Develop open autonomous payments standards: Create and adopt open, interoperable standards for AI agent identity, authentication, and behavioural authorisation across markets and platforms.
2. Embrace orchestration: Develop a common orchestration layer or enable integration with frameworks that bridge AI agents and existing payment rails without full-scale infrastructure rebuilds.
3. Proactive regulatory alignment: Engage proactively with regulators to shape forward-thinking policies that ensure safe innovation and avoid reactive restrictions.
4. Establish shared governance: Implement standard governance models for agent registration, certification, and accountability to prevent ecosystem fragmentation.