One sentence to a guest. Six facts to an assistant.
Restaurant information changes during service, and each part of a guest's question lives in a different system, some in nobody's. This brief treats it as three kinds of truth, looks at where each comes apart, and marks where an assistant has to hand the question to a person.
CoversRestaurant groups, multi-location operators, breweries, hospitality groups and independents
01 · What the industry knows
Three kinds of restaurant truth.
- Location truth
- Each location's address, hours, holiday hours, phone, parking, accessibility, and which services it runs: dinner, brunch, the bar, private dining, catering.
- Menu truth
- What each location serves, during which service, with descriptions, options, published prices and the dietary statements the kitchen makes.
- Operational truth
- What changes during service: a dish sold out tonight, the tables open at 7:30.
- Facts in nobody's system
- Private dining capacity, parking, the step at the entrance, which location takes large parties.
- Breweries and brewpubs
- Taproom hours and kitchen hours are different facts. So are the tap list, the food menu, and whether reservations are taken.
- Groups
- Every location multiplies the copies, and a group can be right at one door and wrong at the next. Hospitality groups add several concepts under one roof.
Where those facts live today.
| Fact | Where it lives today |
|---|---|
| The location | The restaurant's site, its Google profile, every ordering and reservation listing |
| The menu | The POS or menu provider, a PDF on the site, each delivery app |
| The diet | Menu marks, a footnote, the staff |
| The table | The reservation system |
| The details | A line on the site, a Google attribute, a manager's head |
| The next step | The reservation page the restaurant chose, or the phone |
Each copy is edited separately, and a holiday change reaches some of them. A new seasonal menu goes into the POS today; the PDF, the Google menu and the delivery apps catch up when someone remembers. When the kitchen runs out, the POS knows, the ordering app may know, and nothing else does.
03 · What is being decided
What the guest's assistant is trying to settle.
A menu says what a kitchen makes. An assistant acting for a guest also has to know which location, which service, whether the dish is still offered tonight, whether the restaurant is open at that hour, whether it can take the party or the diet, and what the next step is.
- Future scenario
“Find an Italian restaurant near me with a table for four at 7:30 tonight, gluten-free options and parking.”
| Stage | The assistant's job | What has to be true |
|---|---|---|
| Discover | Find Italian restaurants nearby that are open tonight | Cuisine, locations and tonight's hours are stated and agree everywhere |
| Understand | Read the dinner menu and the dishes marked gluten-free | The menu is in text, by location and service, with the kitchen's dietary statements |
| Qualify | Check a table for four around 7:30, and parking | Availability comes from the reservation system; unknown stays “ask,” never “full” |
| Contact | Find how this location takes reservations | The restaurant's own reservation path, with its party-size and cancellation terms |
| Act | Book, or ask about a larger party | The guest books through the restaurant's system; a private dining request reaches a person |
What we've seen, what we expect, what could come later.
- Observed
In one Lehvel audit of a New Mexico brewery, eight sources showed four different sets of hours and two street addresses, and an AI assistant asked about the town answered with Carlsbad, California.
Source: Lehvel audit.
- Observed
Google began rolling out agentic restaurant booking in AI Mode across the U.S., completing bookings through partners such as OpenTable, Resy and Tock.
Source: Google, The Keyword blog, 2025-11-17.
- Hypothesis
Booking assistants route through the reservation and ordering platforms restaurants already use, and none of them reads a restaurant's own statement of parking, private dining, dietary policy or holiday exceptions as a governed source.
- Hypothesis
When an assistant finds conflicting facts about a restaurant, it picks one, and not reliably the restaurant's own. We haven't measured how often.
- Future scenario
An assistant books the table, joins a waitlist, places an order or takes a deposit through the restaurant's own systems, with the operator's opt-in.
05 · Current Lehvel work
The work behind this brief, with its status.
Not built for this sector: a restaurant record by location and service, read-only answers for assistants, or a private dining and large-party request path.
Repeat the kitchen's words. Hand the allergy question to a person.
- Service facts with a time
- Sold-out items and open tables only from a current reading of the restaurant's own systems. An old reading becomes “ask the restaurant,” and a missing one is never “full.”
- Dietary statements
- An assistant may repeat what the restaurant states about a dish, in its words, with the kitchen's caveat. It never says a dish is safe for an allergy, and never treats silence as “free of.”
- The restaurant's own path
- Reservation and ordering links are the ones the operator chooses.
- Public, on request, internal
- Each fact carries one of three marks, and an assistant gets only what was approved.
Authority boundaryReservations, deposits, orders and payments stay with the systems the restaurant already runs. Whether a dish is safe for someone is the staff's call, every time.
For one businessBusiness Truth Discovery · $750