Yelp now lets an AI agent place calls on a user's behalf to check wait times at restaurants, a small feature with a large implication. Until now, the last mile of local search ended at a phone number, because live availability lived in a conversation rather than on a page. An agent that can dial closes that gap, and in doing so changes what counts as machine-readable information about a business.
An assistant that dials instead of searches
Search assumes the answer already exists somewhere and only needs to be retrieved. Calling assumes the answer has to be produced in real time by a person. Most local commerce sits in the second category: table availability, service queue length, whether a specific item is in stock, whether the kitchen is still serving. None of it is reliably published.
An agent that dials converts those conversational answers into structured data, at least for the duration of the request. That is the meaningful shift. Once a system can obtain live state on demand, the value of stale published information falls, and the value of being reachable and legible to an automated caller rises.
Why phone calls were the last gap
Live inventory and wait times
Live state is the hardest thing to publish because it changes hourly. A restaurant can post opening hours, but not the current wait. A store can list a product, but not guarantee the last unit is on the shelf. Agents that call bridge this gap without requiring every small business to run a real-time inventory system.
Hours and exceptions
Hours are a deceptively large problem. Holiday schedules, private events, kitchen closing times and seasonal closures create a long tail of exceptions that directories handle poorly. Errors here produce the worst possible customer experience, a wasted trip, which is why live confirmation has outsized value relative to its technical difficulty.
Language and local nuance
Automated calling also has to handle local nuance, from dialect to how a host describes a wait. Systems that work well in one market often fail in another, which is why this capability tends to roll out city by city rather than globally. The constraint is not model quality but conversational coverage of local phrasing.
What this means for machine-readable store data
For businesses, the practical lesson is that being findable is no longer enough. An agent needs structured, unambiguous facts: exact address, service area, hours with exceptions, price range, and a phone line that a human actually answers. A listing that is accurate but unreachable becomes a dead end in an agent-driven workflow.
The second lesson is consistency across surfaces. If a store's hours on its own site differ from its directory listing, an agent will pick one and act on it, and the customer will discover the error at the door. Maintaining a single source of truth for location data is unglamorous work that now has a direct revenue consequence.
Brands that will be quoted and brands that will be skipped
In answer-driven interfaces, brands are either cited or absent. Being cited requires that the model can identify the entity, associate it with the right category and attributes, and find corroborating sources. Brands with fragmented naming, inconsistent category descriptions or thin third-party coverage get skipped, not because they are worse, but because they are harder to describe.
The fix is structural rather than promotional. Publish a canonical description of the business, keep the name and category consistent everywhere, and make sure third-party sources repeat the same facts. That is the same discipline that determines whether a brand appears in a generated recommendation, and it is measurable: query the major assistants monthly and log whether the entity is recognised correctly.
A checklist before agents start calling
Five items are worth checking now: a single canonical address and phone number, hours with an explicit exceptions page, category and service descriptions written in plain language, a published price range, and a phone line with a stated response commitment. Each item is cheap to fix and each one removes a reason for an agent to fail.
The strategic point is that agent-mediated discovery rewards operational clarity over marketing volume. A business that answers the phone, keeps its listing accurate and describes itself consistently will be recommended more often than a louder competitor that does none of those things. That is a shift worth preparing for before the calls become routine.
