Meet Tsuno at Sirha Lyon 2027 · Stand S.57Learn more
EN FR
Try the demo
Menu

Team and management

How AI helps a manager plan their team

AI helps with scheduling by bringing the rota, validated skills and actual hours together. It suggests, flags and drafts. The manager decides and signs off.

Published on

In a restaurant's back office, a manager on the phone looks at the week's schedule on a screen, on a Saturday morning

AI helps a manager schedule better when it brings together what is usually scattered: the week's rota, each person's validated skills, the hours actually worked and the rest rules. It spots an uncovered shift, suggests cover from people who can actually run the station, flags a rule that would be broken and drafts the message to the team. It does not choose on personal criteria, keeps employment law in view, and always leaves the decision to the manager.

Saturday, 7:50 a.m., in the back office

Manon opens the restaurant at 10. At 7:42, Julien texted: he is sick and will not be in. He was down for dish and cold prep from 9 to 3. Saturday lunch is the busiest service of the week.

Manon has the schedule up on the office screen. She knows her team. She knows Aïcha closed late last night, that Pierre is a 17-year-old apprentice, that someone else has already put in a lot of hours this week. But she has to piece it all together in her head, on the phone, while trying to remember who has actually been trained on cold prep.

This is where a useful tool saves ten minutes and prevents a mistake. It does not pick who comes in. It lays out the facts: who is signed off for the station, who is already scheduled, what time each person finished yesterday, how many hours they have done. Manon makes her calls, then updates the schedule.

The scene is fictional. It looks a lot like many Saturday mornings.

Why scheduling a restaurant is still so hard

The information lives in different places

The rota is in one tool, skills are in the head chef's head or on a whiteboard, actual hours are in the time clock, contracts are in a binder. To cover a shift, you need all four. Managers do it from memory, usually in a rush, and memory has gaps.

The planned schedule is not the real one

A close that runs forty minutes over, an extra pair of hands added at the last minute, a skipped break. Week after week, the gap between planned and worked hours grows and nobody sees it. The same schedule gets copied forward, with the same holes.

Employment law cannot be checked by eye

Last-minute cover is exactly when rest rules get broken. In France, the Labour Code requires at least eleven consecutive hours of daily rest (article L3131-1) and a weekly rest of at least twenty-four consecutive hours on top of those eleven (article L3132-2). For young workers, daily rest is at least twelve hours, fourteen under sixteen (article L3164-1), and night work, between 10 p.m. and 6 a.m. for 16 to 18-year-olds (article L3163-1), is banned unless an exceptional exemption is granted (article L3163-2). Calling Aïcha in at 9 after she finished at 12:10 a.m. is already too short. Collective agreements can add specific rules, and if you operate outside France, your local rules are the ones that count.

The hours budget is often arbitrary

Many schedules are built against an hours figure set once and never revisited. We explain why that misleads in the trap of an arbitrary labour-hour budget. A useful budget starts from expected trade and the stations that need covering.

What AI does well in scheduling

  1. Flag an uncovered shift. A recorded absence leaves a gap on a specific slot and station. The tool says so straight away, with the time and the station, instead of letting the manager find out at opening.
  2. Suggest cover from people with the skill. Not just anyone who is free: people whose skill for that station has been validated and dated by a manager. The list is short, and each name comes with the reason it is there.
  3. Flag a rest or hours rule. "Aïcha finished at 12:10 a.m.: bringing her in before 11:10 would cut her daily rest below eleven hours." "Pierre is 17: no shift after 10 p.m. without an exemption." The reminder is factual, not an automatic block.
  4. Show planned versus actual. Last week, which station ran over, which hours piled up on the same people, which contracts are exceeded or underused.
  5. Draft the message. The note to the team or to the person being asked, short and specific. The manager reads it, edits it and sends it.

What it should not do

  1. Decide for the manager. Manon knows Pierre has exams this week, and that a server asked not to close for a reason Manon has no business writing down. The tool does not know everything, and that is fine. It suggests; she decides.
  2. Choose on personal criteria. Picking someone because they "always say yes," or leaving out someone seen as "unreliable," is not a tool's job. The AI does not score or rank employees. It works from work facts: skills, hours, contract, availability.
  3. Ignore employment law. A tool that quietly suggests a shift that breaks daily rest, or puts a minor on a night shift, is worse than useless: it just makes the mistake faster.
  4. Change the schedule on its own. Nothing goes out to the team until the manager has seen it and confirmed it.

The Saturday morning table

SituationWhat the AI suggestsWhat the manager decides
Julien off sick, dish and cold prep 9 to 3Flags the uncovered station, lists three people trained on cold prep who are not scheduledWho to call first, or whether to reshuffle stations among those already in
Aïcha finished at 12:10 a.m.Flags that bringing her in before 11:10 would cut her rest below eleven hoursNot to ask her, or move her start if she agrees and the rules allow it
Pierre, a 17-year-old apprentice, offers to come inFlags the twelve-hour rest and the night-work banTake him for the lunch shift, within those rules
Someone already has a lot of hours this weekShows their hours balance and the gap with their contractWhether to ask them, and how to make it up later
The same station runs over every SaturdayShows planned versus actual for the last four SaturdaysRework the standard Saturday schedule or its hours budget
The schedule changesDrafts the message to the people affectedRead, edit, send

The third column cannot be delegated. It holds what the manager knows and the tool does not, such as an arrangement made with an employee or how tired someone was the night before.

Common mistakes

  • Assuming free means available. Without a validated skill for the station, the cover fills the slot but not the service.
  • Fixing the absence and forgetting the week. Saturday cover moves hours around. If nobody checks the balance, the problem is back by Thursday.
  • Leaving skills in the head chef's head. When the chef leaves, nobody knows who can do what. That is what building a skills matrix that actually works is about.
  • Asking the tool to pick "the best one." There is no best employee in the abstract. There is someone who has the skill, can come in without breaking a rule, and a manager who chooses.
  • Sending the message unread. A drafted message is still a draft.

What to measure

  • Time from absence reported to shift covered, over your last five last-minute absences.
  • Gap between scheduled and worked hours, by station and by week, over a month.
  • Number of cover shifts that broke a rest rule, checked after the fact. The target is none.

Where Tsuno fits

In Tsuno, the weekly schedule is built per site, copied from one week to the next and published to the team by email and push notification. Every employee sees their shifts on their phone. The time clock on the site tablet records hours worked, the hours balance shows the running total and this week's gap, and gaps between contracted and scheduled hours are flagged. Once the POS is connected, an hours budget worked out from expected revenue is shown while you build the schedule. An absence is recorded in one tap from the Today page. The details are on the Features page.

Skills are validated and dated by a manager, and the skills matrix shows who can do what, site by site. That is the raw material you need to suggest cover that can actually run the station. For on-the-job training, see Tsuno Compétences.

You can ask questions in plain language, from Tsuno or from the AI assistant you already use, with your own access rights and nothing more. Nothing runs without your confirmation, and every action is logged. Automatically cross-checking an absence, skills and rest rules to propose a cover list is what a good scheduling tool should do; we would rather show you in a demo than promise it here. Tsuno brings the facts together. People decide.

To see where your labour cost sits against your revenue, try the labour ratio calculator.

Further reading

This article is part of a series on AI for frontline managers: the overview is in AI for frontline managers: analyse, advise, reassure. To look further ahead than next week, read anticipate rather than react: AI and restaurant planning. On hours budgets, see the trap of an arbitrary labour-hour budget. And on where a tool must stop, AI should not judge your employees. Planning also means handling the unexpected: see swapping a shift without calling the boss, following up deadlines and reminders without a spreadsheet and the pre-service briefing in the age of AI.

Key takeaways

A good scheduling tool puts in front of the manager what they used to rebuild from memory: the uncovered shift, the people with the skill, the hours and the rest rules. It suggests and drafts; the manager knows what the tool does not, and signs off.

Frequently asked questions

Can AI build a restaurant schedule instead of the manager?

It can prepare the ground: spot an uncovered shift, suggest people who are signed off for the station, flag a rest rule, draft the message to the team. It should not decide. The manager knows things the software does not, and signs off every change.

How do you find cover quickly for a Saturday morning call-in sick?

You need three things in one place: who is already scheduled, who is signed off for the station, and who can still work without going over their hours or cutting into their rest. A good tool cross-checks all three and offers a short list. The manager makes the calls and chooses.

Which rest rules should you check before changing a schedule in France?

French labour law requires at least eleven consecutive hours of daily rest and a weekly rest of at least twenty-four consecutive hours on top of those eleven. Workers under 18 get at least twelve hours of daily rest and night work is banned in principle. Collective agreements can add specific rules. Other countries have their own rules, so check locally.

Why compare the planned schedule with the hours actually worked?

Because the gap tells you what the schedule does not: a station that is always understaffed, a close that runs over, hours piling up on the same people. It is the basis for fixing next week and keeping labour cost under control.

Can AI decide who works based on how reliable or motivated people are?

No. A scheduling tool should not score people or decide on personal criteria. It works from facts: validated skills, hours, contract, stated availability. Choosing between two people stays with the manager.