Merchants are currently integrating AI shopping agents into their existing sales infrastructure rather than replacing current payment stacks. According to Pavan HS, senior director of product for agentic commerce at Visa, businesses must ensure these agents function within established systems for payments, fraud detection, loyalty programs, and returns.
Technical challenges include managing product data for AI agents, which may access information via web crawling or specific protocols. Merchants must decide whether to sell through third-party AI platforms or develop proprietary chatbots. A key issue remains the correlation of transactions across channels, such as when a purchase made via Google Gemini is returned at a physical store.
Visa utilizes token management services to link credentials across different payment environments. For example, the company employs agentic tokens tied to specific purchasing intents and FIDO-based biometric authentication to verify user identity. These tools help address complications where AI agents alter standard transaction signals, such as originating traffic from different geographic regions.
Identifying legitimate agents versus malicious bots is a significant concern for merchants. Visa has introduced the Trusted Agent Protocol to provide information regarding an agent's origin and intent. Additionally, content delivery networks like Cloudflare and Akamai are developing protocols to evaluate automated traffic before it reaches a merchant's website.
Operational success for merchants depends on ongoing monitoring of authorization rates, chargebacks, and disputes. Integrating AI agents is not a one-time task, as businesses must continuously adjust fraud controls to accommodate new traffic patterns while maintaining security standards for omnichannel commerce.
