Securing Trust in the Age of Agentic Commerce
What happens when an AI agent makes a perfectly valid payment that isn't quite what the customer authorized?
As agents begin to transact on our behalf, banks need to verify more than identity and credentials. They need to know whether each transaction reflects the customer's intent, authority, and boundaries.
See how Infosys is approaching this new trust challenge, creating a verifiable chain from customer intent to payment execution.
Watch the video to see the approach in action.
When a Valid Transaction Breaks the Customer's Rules
In this video, a consumer authorizes an agent to purchase a laptop for no more than $1,500 from approved merchants. The agent completes a $1,600 purchase, creating a difficult choice for the bank: approve a transaction that breaches the customer's boundary, or decline an otherwise legitimate customer journey.
A second scenario shows an agent-initiated transaction for which the credential is valid, but the customer disputes ever providing the intent. The examples highlight a central challenge for agentic commerce: credentials prove access, but do not by themselves prove purpose, limits or authorization context.
A Bank-Centered Control Model
The Infosys solution establishes a cryptographically verifiable, contiguous record across the transaction lifecycle, from customer intent and mandate to payment instruction, authorization, clearing, settlement and posting to the core banking system.
Instead of relying only on binary approval or denial, banks can apply policy controls that allow them to proceed, deny or step up the transaction by seeking clarification from the customer. This helps preserve legitimate journeys while reducing disputes and unauthorized boundary breaches.
End-to-End Evidence for Operations and Dispute Resolution
The demonstration presents a bank operations dashboard, transaction explorer and evidence library. Together, these capabilities capture customer-defined boundaries, policy decisions, cryptographic proof, timestamps, payment-message execution and posting into the core banking platform, with Finacle shown as the reference core.
The approach is designed to work across bank-trusted and third-party agents, and across card networks, real-time payment rails, merchant payments and account-to-account payments. A unified evidence trail gives banks a clearer basis for customer explanations, operational controls and dispute resolution.
Building Trust into Agentic Commerce
As agent-led transactions become more common, banks will need to move beyond confirming that a credential was valid. By securing the chain from intent to execution, they can strengthen customer trust, improve accountability and support safer adoption of agentic commerce.
Request for services
Find out more about how we can help your organization navigate its next. Let us know your areas of interest so that we can serve you better.