Occupancy summaries from POS data

Every week the manager of a restaurant or cafe builds the shift rota from what they remember from last week, who can work and when an event is coming. Lightspeed or Poster POS meanwhile holds the whole history: how many receipts there were each hour, the average check, how many of them were Wolt and Bolt Food orders, which load the kitchen but not the dining room, and how traffic changed on holidays, in rain or during city events. The booking system already shows next week's parties, and the calendar shows the wedding and the company evening. Nobody puts these sources together, so in some hours hands are short and guests wait, in others staff stand idle, and the payroll budget is exceeded.

We combine the hourly sales export from Lightspeed or Poster POS, the booking system, Wolt and Bolt Food order reports and the events calendar. Each week, via API, Claude analyses previous weeks' sales by hour and day, separates dining room receipts from delivery orders, applies the known bookings and events, takes holidays and any exceptional occasions you specify into account, and then prepares next week's occupancy forecast and a suggested staffing need for each hour according to your ratios (for example, how many tables per waiter, how many Wolt orders per hour for the kitchen). The result is delivered in a Power BI report and a short summary: which hours will be busy, why, and how many people are suggested in the dining room, at the bar and in the kitchen. The manager sees the reasoning, builds the rota in their usual system and decides on exceptions. The system assigns no employees, changes no rota and does not communicate with staff. The model runs on the Anthropic API or Amazon Bedrock in an EU region, and employees' personal data is not passed to the model, because the forecast needs only sales and occupancy figures.