The restaurant AI implementation checklist.

Discovery is easy. Test calls are fun. Menu prep is the week that determines whether the rollout succeeds. Set aside real operator time — the menu is the source of truth for everything the agent does.

Two of the checklist items — recording enablement and AI-disclosure copy — should be reviewed by qualified counsel in your jurisdiction. Fire It offers configurable disclosure text; the specific words are your decision, not ours.

Assume a single-location pizzeria starting today. Week zero: pull two weeks of carrier data, compute unanswered percentage by hour, and run those numbers through the missed-call calculator. Budget roughly four operator-hours across the week; anything less and the discovery inputs are guesses. Week one: import the menu, then walk the menu voice-readiness checklist end-to-end. Aim to finish alias coverage for the top 20 items by the end of the week. Week two: configure hours, transfer targets, greeting copy, and disclosure wording (with your counsel), then run six scripted test calls — a plain takeout order, a modifier-heavy order, a "let me talk to a human" transfer, a sold-out substitution, an address edit mid-call, and a call outside business hours. End of week two: flip busy-forward to Fire It only.

Week three is a monitoring week, not a launch week: daily transcript review, one alias tune-up mid-week, and a Friday go/no-go on whether to add no-answer-forward. Most single-location shops that follow this shape land at "overflow-only, ~40% of inbound" by end of week three. Groups with more than one location should still start with a single-shop pilot on this same cadence, then extend using the multi-location playbook.

The three-week estimate assumes an operator with basic POS admin access, a menu that's already structured (not a scanned PDF), and someone willing to spend four focused hours in week zero on carrier data. Multi-location groups, shops migrating from a legacy phone system, or menus that need a full rebuild should plan for four to six weeks instead. The checklist doesn't cover accounting integration or hardware procurement — Fire It runs on your existing carrier and POS, so those steps are deliberately out of scope.

Progress persists in your browser only and can be cleared with the "Reset" control. Two items — recording enablement and disclosure wording — are jurisdiction-specific legal decisions that Fire It does not make for you; see the AI & call disclosure notes and consult qualified counsel before flipping either toggle. State-specific framing questions for your counsel review are collected in the guides at /restaurant-ai/california, /texas, /florida, /new-york, and /illinois.

Week zero looks unglamorous and it's the week that decides whether the rollout works. Pulling the carrier report, estimating average ticket, and identifying the peak-hour cluster take a total of about four operator-hours. Every hour skipped here is repaid tenfold in the tuning phase. Fire It's guidance is to complete discovery before creating the account — the account setup takes fifteen minutes and can wait until you have real numbers to configure against.

The recommended rollout shape is overflow-first: point the busy-forward at Fire It for two weeks, monitor daily, and only then decide whether to expand to unanswered-forward, then to primary. That progression builds confidence gradually and keeps every failure mode small. Shops that flip straight to primary sometimes do fine, but the recovery from an early mistake is much easier at overflow scope than at primary scope. Nothing about the technology requires primary first.

Two checklist items always require your counsel: whether to enable call recording, and the exact wording of your AI-disclosure line. Fire It ships configurable defaults for both, but the defaults are starting points for a counsel review, not compliance guarantees. State laws vary meaningfully — two-party vs one-party consent, industry-specific rules, evolving AI disclosure requirements. Retain qualified counsel for your jurisdiction and don't launch either recording or disclosure copy without their sign-off. This is one of the few places where 'move fast' costs more than it saves.