Today's Edition
SADEQ ALKHOORI (ABU DHABI)
As AI agents move closer to shopping, booking and paying on behalf of consumers, payment companies and merchants are looking at how consent can be verified when a customer is no longer completing every step manually.
The shift could move parts of online commerce from a human-click model to one where software acts within limits set by the customer. For payment networks, banks and merchants, the key issue is how to prove that a transaction reflects what the customer actually approved.
Prakriti Singh, Executive Vice President for Core Payments, EEMEA, at Mastercard, told Aletihad that customer intent becomes harder to see when an AI agent acts independently.
“When an AI agent interprets a consumer request and acts independently to make the purchase, intent becomes less visible … and becomes more important to verify,” Singh said.
She said consumers need confidence that agents will follow instructions; merchants need assurance that an agent is authorised to transact; and issuers must be able to distinguish legitimate activity from fraud.
“Meeting these requirements requires a new model of trust,” Singh told Aletihad.
According to Singh, Mastercard is working with Google, Adyen, Checkout.com, Getnet, Fiserv, and other partners on “verifiable intent”, a system designed to create a tamper-resistant record of what a user authorises an AI agent to do on their behalf.
“If a dispute occurs, all parties can rely on a clear audit trail to resolve it quickly and fairly,” she said.
Singh said AI agents must be uniquely registered, connected to the user they represent, and recognised across the transaction chain. Mastercard has adopted a Know Your Agent (KYA) registration framework, designed to build trust in AI-driven interactions, similar in principle to the Know Your Customer rules used by banks.
She said verified agents are issued digital identifiers, while agentic tokens are used for each payment request to help ensure actions remain traceable, secure, and within the spending rules approved by the user.
“User intent must never be assumed. It must be verified with biometric authentication, consented and central to every interaction,” Singh told Aletihad.
In the UAE, merchant readiness will also shape how quickly AI-agent payments move from early use cases to wider adoption.
Sayed Munaf, Chief Business Officer at Telr, told Aletihad that UAE merchants have shown a strong record of adopting payment innovation, from digital wallets to ‘buy now pay later’ solutions.
“If AI-agent payments can reduce friction, improve customer experience and increase conversion rates, we believe merchants will be receptive,” Munaf said.
Trust Needs New Rules
However, he said adoption would depend on customer trust and the safeguards surrounding each transaction. “The key determinant of adoption, however, will be consumer trust and confidence in the underlying safeguards,” Munaf told Aletihad.
He said merchants would need clear evidence that a customer authorised an AI agent to act on their behalf. He said this would require robust authentication, explicit consent, and auditable transaction records showing the scope of authority granted to the agent and the actions it performed.
“As with any emerging payment technology, agentic commerce introduces new considerations around fraud, disputes and liability,” Munaf said.
He said the industry would need clear frameworks for cases involving unauthorised activity, mistaken purchases, or disputes over customer intent, while ensuring appropriate protections for both consumers and merchants.
For payment gateways, Munaf said the core safeguards remain strong customer authentication, tokenisation, fraud monitoring, transaction controls and comprehensive audit trails.
He said agentic commerce is also likely to require more detailed permission management, spending limits and real-time oversight to ensure customers remain in control of transactions made on their behalf.
Mastercard introduced Mastercard Agent Pay last year as a trust layer for agentic payments. Singh said the solution allows merchants to identify and accept transactions from verified digital agents, using agentic tokens that keep payments within limits defined by the user.
She said Agent Pay blocks payments if a malicious actor tries to take control of an agent acting for a consumer. The solution debuted in the UAE in November 2025 in collaboration with Majid Al Futtaim, marking its first launch outside the US.
Munaf said agentic commerce is moving beyond early discussion: “Agentic Commerce is no longer a theoretical concept. As these capabilities mature, the focus will shift from technical feasibility to trust, governance and consumer protection.”
