Restaurants

Questions Restaurants owners ask first

The same questions come up on most discovery calls. Here are the short answers.

What systems do you need access to before we start, and how do you handle data living across Toast, Square, Resy, OpenTable, and 7shifts?

The audit needs read-only access to whatever is generating the data we're scoring. For most independent operators that means the POS (Toast or Square most often), the reservation platform (Resy or OpenTable), the Google Business Profile and Yelp listings, and the email or SMS channel where catering and event inquiries land. 7shifts comes in if labor scheduling is part of the scope. We don't need merged data on day one. The audit works off whatever exports or API access exists and notes where the gaps are. For a build, we wire the integrations the specific automation needs — review-responder hits Google and Yelp directly, a confirmation cadence pulls reservations from Resy or OpenTable, a catering triage reads the inbox. We don't try to consolidate everything into one platform unless that's the build you're paying for. Restaurants run on five to eight tools and that's fine.

How do you keep AI from saying something stupid about allergens or food safety in a customer reply?

Allergen and food-safety claims never get generated freely. The system is configured to recognize trigger phrases — gluten-free, nut allergy, dairy-free, kosher, halal, undercooked, sick after eating — and route those messages out of the auto-response path entirely. They land in a queue the GM or owner sees within minutes, with a suggested human-written reply drafted but not sent. The auto-responder handles the safe stuff: thank-yous on positive reviews, reservation confirmations, hours and parking questions, generic catering inquiries. Anything that touches health, allergy, or a complaint with bodily-illness language stops at human review. This is non-negotiable on our end. We'd rather slow down a thank-you reply than auto-send a sentence that creates a liability issue. The escalation rules get documented in writing and reviewed with you before launch so there's no ambiguity about what the bot will and won't touch.

A review-responder sounds useful but I worry about it sounding fake. How does that actually work?

The review-responder reads the review, classifies it (five-star generic, five-star with specifics, three or four-star mixed, one or two-star complaint), and either drafts or sends based on the rules you set. Most operators run it in suggest-mode for the first two weeks: it drafts the reply, sends a notification to the GM's phone, and waits for a one-tap approve before posting. After that, the safe categories — five-star thank-yous, simple positive replies — go to auto-send and the rest stay in suggest-mode. Tone matches your existing voice because we feed it your last 20-30 written replies as the style anchor. The tells that make AI replies sound canned (every reply starts the same way, every reply is exactly the same length, every reply ends with the same call to action) get explicitly engineered out. Complaints never auto-send. Those go to you with a draft, the original review, and any matching reservation or POS context the system can pull, so you can decide whether to reply publicly, reach out privately, or both.

When do I see this paying for itself, and what should I expect on no-show rate or review velocity?

Realistic ranges, not promises. A confirmation cadence with one-tap confirm/cancel typically catches a meaningful share of would-be no-shows in time to release the table back to the host or run a short waitlist text. We won't quote you a specific percentage drop without your data — anyone who does on a sales call is making it up. Review-response turnaround moves from "when the owner remembers" to under an hour for the safe categories, often inside ten minutes. That changes what your listing reads like to a diner scrolling Google at 6:30pm trying to pick a place, and review-engagement is one of the signals the platforms weigh on local rank. On catering and event inquiries, the dollar lift is usually the cleanest to measure — a reply that goes out within the hour books at a materially higher rate than one that goes out the next morning, and you'll see that show up in the catering revenue line within a quarter. We set up the measurement before we set up the automation so you can see the lift, or the lack of it, in your own numbers.