Navigating the future: how agentic commerce is redefining payments

Dr. Stephen Whitehouse, Oliver Wyman: Agentic AI is transforming financial services by autonomously handling commerce – from discovery to payment – forcing banks and ecosystem players to rethink roles, risk, infrastructure, and value capture.

 

The financial services industry is at a pivotal moment as autonomous, agentic AI begins to take over tasks traditionally handled by humans. Major players such as OpenAI, Google, Microsoft, and Anthropic are leading this transformation by integrating AI agents into ecommerce to autonomously handle everything from product discovery to price negotiation and payment processing. As this technology evolves, financial institutions must confront key questions: How will agentic AI redefine their role in the ecosystem? What new risks will emerge? And how can banks capitalise on this shift? This article explores how agentic commerce is reshaping the payments landscape and offers insights for banks to stay ahead.

How the industry is adopting agentic AI

AI companies are working with financial institutions and ecommerce players such as Stripe, Mastercard, Klarna, Etsy, and Shopee, as well as crypto wallet providers Coinbase and MetaMask, to create the new reality. They develop standards and protocols that both enable agentic commerce and help manage its risks.

These standards are emerging across three layers that mirror the stages of commerce.

1. The discovery layer defines product and catalogue standards optimised for agents, making it easier to search, compare, and evaluate offers.
2. The checkout layer sets technical and operational standards from product selection and cart configuration through purchase finalisation and consent capture.
3. The payment layer establishes how to execute and process agent-initiated payments using existing methods and rails in a safe, predictable way.

Although these protocols cut across layers, two initiatives are especially important today at checkout. The Agentic Payment Protocol (AP2), initiated by Google and supported by about 60 financial and ecommerce companies, and the Agentic Commerce Protocol (ACP), created by OpenAI and Stripe, both aim to define how consumers instruct AI agents to search, select, and purchase products with minimal human involvement and how those agents interact with merchant systems and payment providers.

At the payment layer, schemes such as Mastercard and Visa contribute by setting standards on how to execute and process agent-initiated payments safely and securely. As described in Mastercard’s Agent Pay, this includes work on agent-specific tokenised credentials to protect consumers and counterparties, the authentication of agent identities and permissions, and adaptations to dispute and liability models to reflect delegated decision‑making.

How agentic commerce affects the roles in the ecosystem

The emergence of agentic commerce and these protocols is beginning to affect every part of the payment ecosystem, including the existing roles and responsibilities.

Merchants
Agentic commerce shifts decisions from human consumers to AI agents optimising for price, product fit, delivery, and sustainability. Merchants need to be ‘agent-ready’, with structured product data, real-time inventory and fulfilment signals, and machine-readable trust markers.

Issuers
Cards and accounts will be used increasingly through delegated AI agents rather than direct human initiation. Issuers must securely embed credentials in agent environments, support agent-specific tokenisation, and adapt authentication and authorisation to reflect explicit delegation and observable agent behaviour.

Payment service providers (PSP)/acquirers

PSPs/acquirers will move from simple routing to orchestrating agent-initiated, constraint-driven transaction flows. They will need to identify agent-originated payments, verify compliance with spend and policy rules, and upgrade fraud controls to detect anomalous agent behaviour.

Networks
For networks, core authorisation, clearing, and settlement will remain, but rules and frameworks must evolve for delegated, agent-driven transactions. Networks will need to adjust dispute and liability models, support agent-specific tokenised credentials, and align scheme standards with protocols such as AP2 and ACP, keeping them central as trusted infrastructure.

What risks and challenges are raised by agent-initiated payments

Liability and fraud risks of agent-initiated payments

As roles across the ecosystem evolve, a deeper organisational challenge becomes clear. Current payment processes and controls were built for humans, not AI agents. Authorisation flows, dispute handling, and risk models all assume that a person decides to pay and can later be consulted or challenged. As consumers authorise AI agents to act on their behalf, this assumption no longer holds.

Concepts of identity, consent, and liability come under pressure when a transaction is technically agent-initiated but economically and legally tied to a person or organisation. Three areas stand out where financial institutions need to rethink their frameworks:

1. Delegated authorisation. Most authorisation journeys still hinge on direct human approval at key moments. Institutions need clear, auditable models for how customers delegate authority to agents, what those agents are allowed to do, and how that delegation can be limited or revoked.
2. Authenticating agents as well as humans. Mechanisms such as 3D Secure (3DS) and Strong Customer Authentication (SCA) confirm who the human is, but not which agent acted on their behalf. Institutions must be able to register and authenticate agents, understand where and how they can transact, and maintain audit trails that stand up to regulatory and legal scrutiny.
3. Updating Know Your Customer (KYC), Anti‑Money Laundering (AML), fraud, and dispute models. ‘Know your customer’ will need to be complemented by ‘know your agent.’ KYC, AML, fraud, and dispute handling must take into account how an agent was configured and instructed at the time of the transaction and use this information when allocating liability across the consumer, agent provider, bank, network, and merchant.

No single institution can fix this alone. Industry-level collaboration is needed to adapt authentication, authorisation, KYC/AML, and dispute frameworks for an agentic world.

Technical standards in a machine-led industry

Real‑time payments, API‑first architectures, and cloud‑based cores have created fertile ground for agentic AI. But agent‑initiated payments demand speed, automated decision‑making, and reliable orchestration that many legacy environments cannot support.

AI agents operate 24/7, execute microtransactions and large volumes continuously, and expect real‑time responses from the underlying infrastructure. Yet much of the industry still relies on batch processing, settlement windows, end‑of‑day cut‑offs, and manual exception handling – approaches fundamentally at odds with an always‑on, event‑driven environment.

To remain part of the value chain rather than sitting behind it, financial institutions will need to act on three fronts.

• Provide standardised, robust APIs that agents can rely on for payment initiation and execution, aligned with emerging protocols.
• Invest in real-time data and logging so agent actions, permissions, and outcomes can be monitored, explained, and audited.
• Modernise orchestration, risk, and compliance engines to operate in near real time, ingesting continuous streams of agent and transaction data rather than isolated payment events.

Institutions that act now can shape how agents interact with their products, data, and infrastructure and capture the associated growth.

How value will move across the agentic commerce value chain

Agentic commerce changes where value is created and who captures it. Agent providers that sit between customers and merchants will take a growing share of economics, while traditional payment revenues come under pressure.

Discovery

In agentic commerce, a growing share of value is created in agent interfaces and AI‑browsers, where consumers and businesses set preferences and interact with their agents while discovering. These environments generate richer, continuous data on intent, behaviour, and context than traditional channels.

As outlined in Oliver Wyman’s report with Google, AI’s Next Frontier for Financial Services Leaders, three models describe how value is created:

1. Assistance economy – agents orchestrate end-to-end journeys based on detailed insight and customer needs, shifting marketing from targeting consumers to being selected by the agent.
2. Adaptive customer experience – data-rich insights enable products and services tailored to their unique preferences, strengthening loyalty, improving cross-sell, and lowering customer acquisition costs.
3. Agentic twins – persistent, configured agents evolve into ‘agentic twins’: personalised digital representatives of the customer that can be addressed directly with tailored propositions, enabling hyper-personalisation.

Checkout

At checkout, value is beginning to concentrate with those who control the point of conversion. OpenAI’s plan to charge Shopify merchants a 4% fee when checkout occurs through the ChatGPT channel illustrates how AI platforms will monetise this position.

AI agents will not be tied to a single rail. They will route across cards, account‑to‑account, and alternative methods based on cost, speed, risk, and user preferences, making fees and economics far more transparent. As a result, traditional transaction margins for banks and PSPs will be squeezed.

Financial institutions will have to participate in agent‑initiated checkout flows to stay visible, while shifting their business models away from pure volume‑based transaction fees toward monetizing orchestration, data, risk, and control services.

Payment execution

In the execution layer, issuers, acquirers/PSPs, and networks can still capture meaningful value by providing a trusted, agent‑aware payment infrastructure.

Growth opportunities will centre on services such as agent‑specific tokenisation of credentials, authentication of both consumers and agents, and advanced authorisation and risk controls tuned to agent behaviour and configuration. Institutions that embed these capabilities into emerging standards and protocols can position them as differentiated security and trust services, rather than undifferentiated utilities.

Conclusion: capitalising on the agentic commerce revolution

Agentic commerce is transforming the financial landscape, presenting both challenges and opportunities for banks. To thrive in this new world, financial institutions must embrace emerging standards, modernise their infrastructure, and redefine their value propositions. By acting now, they can mitigate risks, capture value from AI-driven processes, and position themselves as leaders in the evolving payment ecosystem. The rise of agentic commerce is inevitable, and banks that adapt quickly will shape the future of financial services.

This article is part of the The Paypers` Money Movement in 2026: Trends in AI, Payments & Regulation Newsletter, a source of expert insights on the forces reshaping fintech, payments, and banking, covering fraud and financial crime, AI in fraud prevention and risk intelligence, real-time and cross-border payments, European payments sovereignty and the future of instant rails, stablecoins, agentic commerce, compliance and the evolving regulatory landscape, payments fragmentation driven by geopolitics and regulation, infrastructure bottlenecks in banking modernisation, the shift from generic scale to verticalised, value-added payment models, and digital wallets changing consumer payment behaviour – delivering a clear, data-backed view of what will shape strategy and innovation in 2026 and beyond.

Explore the other contributions in this Money Movement in 2026: Trends in AI, Payments & Regulation Newsletter series for more expert insights:

• Bizum - European instant payments: sovereignty and cross-border interoperability
• McKinsey - Payments in transition: the forces shaping the future of global money movement
• Triple-A - Stablecoins, the once-in-a-lifetime shift in payments
• UniCredit - Navigating regulatory change: compliance and the future of digital payments
• Unlimit - Why the future of digital wallets depends on interoperability at scale
• ACI Worldwide - Global orchestration for a real‑time, APM‑driven payments era
• The Paypers - 2026 fraud forecast: AI, deepfakes, and rising cybercrime risks

About the author

Stephen Whitehouse leads Oliver Wyman’s Digital Economy & Payments practice in Europe. He advises CEOs, boards, and senior executives on customer, digital, and business transformations, product innovation, and large-scale changes across banks, retailers, fintechs, and regulators. As a published thought leader and regular voice of the industry, Stephen combines industry expertise with hands-on delivery to drive growth, reduce risk and scale new business models.

 

About the company

At Oliver Wyman, a Marsh (NYSE: MRSH) business, we bring deep industry insight, bold innovation, and a collaborative approach that cuts through complexity to help organisations navigate their most defining transformative moments.