Draft next week's schedule against your sales forecast and labor target, from ChatGPT or Claude

Building next week usually starts the same way: open last week, copy it, then stare at it trying to decide whether it's too heavy for a slow Tuesday or too thin for a Saturday. The sales numbers are in one report, the labor target is in your head or a spreadsheet, and the schedule is somewhere else.

With Mays connected to ChatGPT or Claude, you can ask for a draft in one message and get it with the forecast and the labor math next to it, from the same data Mays uses everywhere else.

This post is part of our series Run your restaurant with your AI.

The situation

You run Riverside, one of Harbor Lane Café's two locations. It's Thursday, next week isn't built yet, and you want to stay around 26% labor.

Before: you copy last week in the schedule, export last month's sales to a spreadsheet, average the Tuesdays by hand, guess the labor percent, and paste numbers back and forth between the spreadsheet, the schedule and maybe a chat window to sanity-check it.

Now:

Suggest next week's schedule at Riverside. I want to stay at 26% labor, and there should always be at least 2 people on.

What the AI does in Mays

1. It finds Riverside among your locations.
2. It asks Mays for a suggestion for next week (suggest_schedule). This is a preview: nothing changes. Mays runs the same auto-scheduler as the Auto-schedule button on the schedule. It keeps every shift already in the week and proposes what to add: last week's pattern carried forward, checked against approved time off, availability, the notes you keep on each person, minors' work rules and the weekly hours cap.
3. It shows the forecast next to it. The sales forecast by day and daypart, the projected hours and, if your connection includes pay, the labor cost and labor % against the 26% you asked for (or the location's own target if you don't name one). With "at least 2 people on", it also flags the dayparts where fewer would be on at once.
4. You decide. If it looks right, you say "Add it." The AI asks Mays to add the suggested shifts (apply_suggested_schedule). Mays sends another preview first: how the week's hours change, coverage gaps it closes, and anyone it would push into overtime. Your yes adds them, as drafts.

Drafts are the important part. Nobody is notified and your team sees nothing until you publish the week, the same as when you build it by hand. Adding a suggestion never removes a shift and never publishes.

What you see

An answer in words, for example: "Riverside, next week: 18 suggested shifts on top of the 6 already there. Forecast is highest Saturday lunch. Projected labor is 27.4% against your 26% target, so about 9 hours over. Tuesday 2–5pm would have 1 person on." The numbers are illustrative; yours come from your own sales and schedule.

From there you keep talking: "Drop one of the Tuesday afternoon shifts", "Who would the Saturday open go to?", "Put it on the schedule."

How to read the forecast

Mays's forecast averages the same weekday over the last 8 weeks, and skips days with no sales. That makes it run a little high (a closed holiday doesn't drag the average down), so labor % measured against it tends to read a little low. Treat a result just under target as "about on target".

A location with no sales history in Mays yet has no forecast, so there is no labor % to compare against. Sales reach Mays from your connected POS.

What Mays doesn't do here

• It doesn't publish. The suggestion becomes drafts; publishing is a separate step you ask for when you're ready.
• It doesn't take half a suggestion. "Add it" adds every suggested shift for that location and week. If you only want some, ask for those shifts one at a time instead, or add them all and remove the ones you don't want: removing a draft is immediate and nobody is told.
• It doesn't block anything. Overlaps, time off, availability and overtime are warnings. The AI tells you; you decide.
• It doesn't show money without pay access. Without Include wages, labor cost and tips, you see hours and the forecast, but no labor cost or labor %.

Before you start

• Connect Mays to your AI app. mays.app/ai has the setup for ChatGPT and Claude; Connecting Mays to your AI app has the details.
• A suggestion is a read. Asking for one works on any connection.
• Adding it is a change. That needs an admin to turn on Allow AI apps to take actions (Settings → Connected AI apps → Organization) and you to tick make changes when you connect. Every change is a preview first; nothing happens until you say yes.

If you'd rather tune the target first, see labor targets in plain words. And once the week is drafted, stop overtime before it happens.

Try Mays free for 30 days, with no credit card, and connect it to the AI you already use.