Visa and Mastercard are moving further into agentic commerce as payment networks prepare for a world where AI agents can search, choose and pay on behalf of consumers and businesses.

Latest announcements suggest agentic commerce is moving beyond product discovery and into payment infrastructure. Visa has unveiled new capabilities covering AI-led payments, token assurance, merchant readiness and stablecoin settlement. Mastercard has introduced Agent Pay for Machines, a service designed for machine-driven transactions across cards, accounts and stablecoins.

But as AI agents are prepared for checkout.  recent research from Checkout.com, PSE Consulting and Ecommpay suggests the payments industry still has to solve a harder problem before agent-led shopping can scale. Consumers may want convenience, but many still want clear permissioning, visible controls, trusted payment flows and a way to challenge transactions when something goes wrong.

Networks move towards agent-led payments

Visa used its Payments Forum in San Francisco to announce several AI, stablecoin and token capabilities aimed at what it calls intelligent and programmable commerce.

Its Visa Intelligent Commerce platform is designed to support agentic commerce, where AI agents can securely discover, initiate and complete transactions within defined user permissions.

The company also unveiled a strategic collaboration with OpenAI to allow AI agents to initiate Visa payments within set user controls. OpenAI will provide the conversational interface, while Visa will provide the underlying payment infrastructure.

Other Visa tools include Agent Score, which lets merchants assess whether AI agents can navigate and complete tasks on their websites, and Agentic Directory, which will help merchants and agents verify trusted participants.

Visa has also introduced a Large Transaction Model, trained on billions of transactions, to improve fraud detection and authorisation performance while reducing false declines.

Jack Forestell, chief product and strategy officer at Visa, said: “AI is transforming the front end of commerce. Stablecoins are reshaping the back end. Visa’s role is to enable it to work securely, reliably and at global scale, for every participant in the ecosystem.”

Mastercard’s latest announcement focuses on machine-led payments. Agent Pay for Machines builds on its wider Agent Pay programme and is designed for high-frequency, low-latency and low-value transactions carried out by agents or machines.

The service supports credentialing, controls and guaranteed multi-rail settlement across cards, accounts and stablecoins. More than 30 companies are among the initial participants and supporters, including Adyen, Ant International, BVNK, Checkout.com, Cloudflare, Coinbase, Global Payments, OKX, Stripe and Tempo.

Together, the Visa and Mastercard announcements suggest agentic commerce is moving quickly from an ecommerce concept into payment network design.

Consumers want help, but not blind delegation

Consumer research points to interest in AI shopping, but also clear limits around trust.

Checkout.com research says nearly a quarter of UK consumers expect at least 10% of purchases to be AI-driven within a year. At the same time, 41% of UK consumers said they trust no organisation to operate an AI shopping agent, and 37% said they would never delegate purchases to AI.

Control appears to be one of the main conditions for adoption. UK consumers told Checkout.com they would need features such as easy cancellation, instant permission withdrawal and the ability to see options before a purchase is made.

Ecommpay’s research tells a similar story. Its white paper says that 59.2% of consumers had heard of using agents for online purchases and 73.4% thought agentic shopping would become common. Yet 50.1% said they would not trust an agent with their card details.

The gap is especially visible at checkout. Ecommpay found that 53.3% of consumers use AI agents to compare prices, while 47.1% use them to find discounts or better deals. Only 13.9% said they wanted agents to complete purchases with their approval.

That suggests shoppers may welcome AI earlier in the buying journey, while remaining more cautious once payment credentials and spending authority are involved.

Advertising adds another trust question

PSE Consulting’s research adds a further question around how AI shopping tools will make money.

Its survey of 4,250 consumers across the UK, US, France and Germany found that 43% would choose a free AI shopping assistant even if recommendations were influenced by advertising. By comparison, 27% would pay for a fully impartial alternative.

Consumers are not necessarily rejecting commercial influence. Many appear willing to accept a familiar trade-off from search and social media: free access in exchange for advertising.

But trust is still fragile. PSE found that 40% of consumers said advertising would reduce their trust in AI-generated recommendations, rising to 48% in the UK. Older consumers were more cautious, while younger consumers appeared more comfortable with commercially influenced recommendations.

For agentic commerce, this creates a sensitive issue. If an AI assistant recommends a product, chooses between merchants and completes a payment, consumers may want to know whether that decision came from price, suitability, availability, paid placement or another commercial incentive.

Liability moves up the payments agenda

Merchants are also preparing for a different kind of checkout.

Checkout.com found that 61% of UK merchants believe consumers will adopt agent-led shopping faster than most merchants are prepared for. The same research found that 63% believe payment providers will play a decisive role in making agentic commerce safe and scalable.

Liability remains one of the most difficult questions. If a customer disputes an agent-led purchase, 36% of merchants believe the payment provider will be expected to carry the cost.

That raises a set of practical questions for payment service providers, schemes, acquirers and merchants. Who verifies the agent? Who confirms the customer gave permission? Who checks spending limits? Who handles refunds and chargebacks? Who carries responsibility if the agent buys the wrong item or acts outside the customer’s intention?

Scott Dawson, CEO of DECTA, a payment processing and acquiring company, said the development of agentic commerce will involve a wide range of participants across the ecommerce chain.

“As is the case with cross-border payments, where many parties contribute to the transaction journey, agentic commerce will use a complex network of participants,” Dawson said.

He said clear accountability would be needed across the new ecosystem, including transparency around roles, responsibilities and consumer protections.

Dawson also warned that smaller businesses must have access to the same trusted payment infrastructure as larger enterprises.

“The smallest SMEs must have access to the same trusted payment infrastructure, customer reach and growth opportunities as the largest enterprises – agentic commerce should lower barriers to participation, not create new ones, and support all businesses to benefit from the next generation of digital commerce,” he said.