Stop overtime before it happens: a Wednesday check with ChatGPT or Claude

Overtime rarely comes from one bad decision. It comes from a shift picked up on Monday, a close that ran long on Tuesday and a Saturday double nobody re-checked. By the time payroll shows it, the week is over.

The fix is a midweek look at two numbers per person: hours already worked, and hours still on the schedule. With Mays connected to ChatGPT or Claude, that's one question.

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

The situation

It's Wednesday afternoon at Harbor Lane Café. You want to know who's going to cross the line before Sunday, at Midtown and Riverside.

Before: you pull the timecard report into a spreadsheet, copy the rest of the week's shifts from the schedule next to it, add them up per person, remember which state counts daily overtime, and do it again for the second location.

Now:

Who's going into overtime this week?

What the AI does in Mays

It asks Mays for the week's overtime risk (get_overtime_risk). For the current payroll workweek at each location you can see, Mays:

• adds the hours each person has already clocked to the hours they're still scheduled;
• applies the location's own overtime rules: the weekly threshold, and daily overtime and the seventh day where the state has them. It's the same engine behind the overtime warnings on the schedule;
• lists people already in overtime, people the rest of their schedule will push over, and people within a few hours of the weekly threshold (4 by default), worst first, with their hours per location and their next shift.

A typical answer, in the AI's words: "Midtown: Jordan R. is at 31.5 hours with 12 more scheduled, so 3.5 hours over by Saturday. Riverside: nobody over; Maya L. is 2 hours from the threshold." (Illustrative names and numbers.)

Hours always appear. The overtime premium in dollars appears only when you can see pay in Mays and ticked Include wages, labor cost and tips when you connected.

Then move a shift

Give Jordan's Saturday close at Midtown to someone who has room.

The AI looks at the schedule for Saturday and at who's under the threshold, suggests a name, and asks Mays to reassign the shift (update_shift). Before anything changes, Mays sends a preview of what the change does to the week: the hours before and after, coverage gaps it opens or closes, and whether it puts anyone else into overtime. Time off, availability and double-booking come back as warnings, never blocks. You say yes, and the shift moves.

One detail that trips people up: a change to a published shift is staged. Like editing in the Mays schedule, your team sees it, and the people affected are notified, when you publish the week again. So finish with:

Publish the Midtown week.

The publish preview lists who gets notified (by default only the people the change affects) and any double-bookings or overtime left. Your yes sends it.

Two rules worth knowing

• Overtime is counted per location. Hours at Midtown and hours at Riverside aren't added together, because Mays applies each location's rules on its own. Someone who goes over only when you add the two locations together is flagged separately, so you can check how your payroll treats it.
• Salaried people are left out. They're exempt, so they don't appear.

Next week works too: "Who's over next week?" reads the schedule only, since nothing is clocked yet.

What Mays doesn't do here

• It doesn't change shifts on its own. It shows the risk; you decide what to move, and every move has a preview and waits for your yes.
• It doesn't refuse a shift because of overtime. Overtime is a warning everywhere in Mays. Sometimes the extra hour is the right call.
• It doesn't estimate pay it can't see. Without pay access there are hours but no dollar amounts, never a guess.

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.
• The overtime check is a read and works on any connection.
• Moving or publishing a shift is a change. An admin turns on Allow AI apps to take actions under Settings → Connected AI apps → Organization, and you tick make changes when you connect.

To build a week that starts closer to target, see drafting next week against your forecast. For finding a replacement when someone drops a shift, see who can cover Friday's close.

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