AI in hospitality and food service: footfall can be forecast, invoices reconciled.

Covers and occupancy history, supplier invoices, spec sheets, customer reviews, staffing: the data already exists in your tills, your PMS and your inboxes. The question is not to produce more of it, it is to put it to work.

The use cases above read by profile, in your vocabulary. They describe possibilities: what applies to you is decided at the diagnostic, on your real figures.

Forecasting covers or occupancy from history, seasonality, weather and local events is a classic demand-forecasting use case, brought down to the scale of a single establishment.

The possible gain is twofold: orders set on the forecast rather than on habit, so less waste; and staffing sized to expected footfall rather than rolled over.

The forecast proposes, the operator disposes: the chef and the manager keep control of orders and staffing. It is the rule across our whole practice: the machine reads, reconciles and alerts; decisions stay human.

An establishment receives dozens of invoices every week: food, drinks, linen, maintenance. Each one can be reconciled against the purchase order and the delivery note, with quantity or price discrepancies flagged before payment.

The regulatory deadline is coming: from 1 September 2026, every company established in France must be able to receive electronic invoices, under the reform led by the French tax administration (DGFiP); the obligation to issue follows in September 2027 for SMEs. Structured invoices lend themselves to automatic reconciliation.

Replying to every review, in your tone, is real work: a reply can be prepared for each one, then reviewed and posted by you. The rule is absolute: no reply goes out without approval.

For recurring enquiries (opening hours, parking, allergens, availability), a supervised assistant can answer on your website. The setting is known: from 2 August 2026, a conversational agent exposed to the public must flag itself as such, the transparency obligation of the EU AI Act (Regulation (EU) 2024/1689).

And as everywhere, unofficial use is probably already happening: a French Odoxa survey finds that 12% of employees already use AI at work, and nearly one in two have never been trained for it. A one-page policy and company accounts close that grey area.

Everything above describes possibilities; the diagnostic turns them into decisions, on your figures. The five-stage protocol applies: Diagnostic, Scoping & compliance, Pilot, Deployment, Managed operations.

The pilot is chosen narrow and measurable: one invoice flow, the forecast for a single service, the review replies for a single establishment. We are paid for the project, never for reselling licences.