Danske Bank and Mastercard (NYSE: MA) have completed what both companies describe as Denmark’s first payment initiated and finished by an artificial intelligence agent.
The pilot, announced on 21 September 2026, is an early live test of so-called agentic commerce inside a regulated Nordic bank rather than a standalone fintech demonstration.
In the trial, a customer asked an AI assistant to arrange a coffee tasting through Mastercard’s Priceless.com experiences platform.
The agent then selected the session, completed the booking, and settled the charge on its own using a Mastercard issued by Danske Bank.
Mastercard’s Agent Pay rails handled the payment, while PayOS coordinated the full transaction path from instruction to settlement.
Agent Pay, introduced in April 2025, is Mastercard’s infrastructure for letting verified AI systems pay on a cardholder’s behalf.
Purchases are authenticated with Mastercard Payment Passkeys, and the design is meant to keep the customer in control: the person still grants permission, the transaction remains visible, and the issuer can see that an agent, not the cardholder tapping a website, started the payment.
Erik Gutwasser, Mastercard’s Northern Europe division president, said the aim is to make AI-assisted buying simpler without sacrificing transparency or security.
Mark Wraa-Hansen, Danske Bank’s head of personal banking in Denmark, framed the test as part of a broader look at how customers might interact with financial services.
The bank, he said, is examining the technology with an emphasis on usefulness, trust, and customer control, and the Mastercard partnership is meant to show that new automation can sit inside existing security and transparency requirements.
The Danish transaction follows earlier European Agent Pay pilots.
Mastercard completed what it called Europe’s first live AI-agent payment with Santander in March 2026, then a French first with Worldline and Crédit Agricole.
The Danske Bank case matters because it ran through an incumbent issuer’s card and systems, not a sandbox wallet.
That raises practical questions banks now have to answer: how much authority an agent receives, how that authority is limited, how to separate a customer’s intent from the agent’s interpretation of it, and who bears the loss if the agent books the wrong thing.
For Danske Bank the payment experiment sits alongside other AI work.
The lender has been expanding cloud and model infrastructure with Amazon Web Services, including Amazon Bedrock and AgentCore, to move agent-style tools from tests into production.
Separately, it has piloted a Model Context Protocol server so corporate clients can connect their own AI assistants to bank data through Premium APIs.
The card payment is a consumer-facing counterpart: an agent that does not only recommend an experience but pays for it.
Agentic commerce compresses a familiar shopping sequence. Instead of searching, opening a site, entering card details, and confirming, the user states a goal and the agent carries the purchase to completion within agreed limits.
Supporters argue that removes friction.
Critics note that convenience only works if consent, authentication, merchant acceptance, and dispute rules keep pace.
Mastercard’s approach tries to treat the agent as a recognized party on the network, using tokens and passkeys so issuers can distinguish agent-led spend from ordinary card use.
The coffee tasting was a modest use case. Its significance is architectural.
A large Danish bank and a global network showed that an AI agent can book and pay inside live issuer infrastructure while leaving a verifiable consent trail. It now remains to be see if that model scales to everyday retail. It will depend on controls, customer willingness, and how quickly merchants accept agent-initiated checkout.