Google’s latest update to Maps brings agentic food ordering into its Ask Maps experience. Toast and Square are among the first services supporting orders.
Google says Ask Maps can now handle more complex requests, including food ordering. The dedicated AI agents will be using information such as restaurant availability and user preferences to help complete the task in the most optimal way. Toast already supports ordering through Google Search and Maps through its Order with Google integration.
That creates a new type of payment interaction. A customer may tell an AI assistant what they want conversationally without selecting the exact eating place. Using descriptive prompts, the agent can potentially select a merchant itself, construct an order and move the transaction toward completion without the customer navigating a traditional checkout page.
The payment industry has been preparing for this shift. Shopify and Google introduced the Universal Commerce Protocol (UCP) in January as an open standard for AI agents to interact with merchants. Shopify describes UCP as infrastructure covering the commerce journey from product discovery through checkout.
The abovementioned technology, however, solves only part of the problem. The harder questions concern authorization, accountability and customer protection.
A traditional online purchase normally has a clear sequence: the customer chooses an item, reviews the price, confirms the order and authorizes payment. An AI agent compresses several of those steps into one interaction. That raises a practical question for payment providers: what happens when the agent makes a mistake?
If an AI agent orders the wrong meal, selects the wrong quantity or misunderstands a customer’s instruction, responsibility can become less obvious. The merchant, AI platform, payment provider and customer may all have different roles in the transaction.
This makes agentic commerce as much a governance challenge as a technology challenge. Payment systems need reliable ways to establish that an instruction came from the customer, determine what the customer authorized and resolve disputes when the agent acts outside that intent.
There are several industry initiatives that address such tricky aspects. For example, the A-Comm Evidence Protocol (AEP) developed by A-Comm Technologies, a fintech company founded by former Visa payments executives, addresses transaction evidence issues by creating a record of every step an AI agent takes during a purchase: discovering a product, forming intent to buy, getting authorization, and completing the transaction. The protocol is open to comments by August 14 before finalization.
The ability to connect an AI agent to a checkout is increasingly becoming standard infrastructure. The competitive question may instead be which providers can make those transactions sufficiently secure, traceable and reversible when something goes wrong.
