Building the weekly staff schedule with AI

How the bot works

Nine steps. The rules, the coverage arithmetic and the fairness checks are ordinary code and cost almost nothing to run; the model is only used to read the messy inputs and explain the draft.

1. It gathers what people told you. Submitted availability, approved time off, standing commitments like a class on Tuesdays, and the requests that arrived as a text to a manager rather than in the app — because that is where half of them arrive.

2. It takes your coverage requirement. How many people you need, in which role or station, for each part of each day. Your numbers, whether they come off a forecast, a booking sheet, last year's week, or a manager's judgment.

3. It drafts the week inside your rules. Only qualified people on a station that needs a qualification. The rest between shifts you set. The maximum consecutive days and hours you set. Anybody's hard constraints, including any you have for staff under eighteen. These are your rules, written down where you can read them.

4. It shows why each shift landed where it did. Every assignment names the rule it satisfies and the constraint it respected: available, qualified, inside the hours, rested, and whose turn it was on the measure you chose. A manager can argue with the reasoning instead of re-deriving it.

5. It names the holes instead of hiding them. A shift with nobody qualified and available is reported as unfilled, with who was closest and what would have to change — a rule bent, a certification added, somebody asked. It does not quietly put an unqualified person on a station to make the grid look finished.

6. It checks the week against your fairness measures. Weekend and closing rotation, how hours are spread, who got their preferred days last time, who has been given the short end two weeks running. You choose the measures; the draft reports how it did on each one.

7. A manager approves it. The draft, the unfilled shifts, the fairness report and anything that needed a rule bent go to a manager as one screen. They edit, approve, or send it back. Nothing is published and nobody is texted before that.

8. Then everybody gets their own shifts. Published to the tool you already use, and sent to each person as their own list rather than a photo of a grid — by text, email or the app they already have. With the notice period you are required to give, which you tell us.

9. Changes after publication are requests, not facts. A swap, a call-out, a late availability change: the bot works out who could cover it inside your rules and your hours, in rule order, and hands a manager the shortlist. The manager decides. Every version of the week is kept, so you can see what changed and who changed it.

What you configure

It can be configured to run in any business. This is what that actually means — the list of things that are yours, not ours.

Your coverage requirement

People per role, per station, per part of the day, per location — and where the number comes from, whether that is a forecast, a booking sheet or a manager's call.

Your scheduling rules

Rest between shifts, maximum consecutive days, maximum hours, who may close and who may open, qualifications and certifications per station, and any rules you hold for staff under eighteen.

Your fairness measures

What a fair week means here: weekend rotation, hours spread, preferred days honored, closing shifts shared. Written down, measured on every draft, and yours to change.

Your time-off policy

Cutoffs for requests, how many people may be off at once per role, blackout weeks, and what happens to a request that arrives after the cutoff.

Your swap and call-out rules

Who may swap with whom, how late a swap is allowed, who has to approve it, and the order in which people are asked to cover an open shift.

Your publication notice

How far ahead the schedule goes out, and the notice period you are required to give — you tell us what applies to you, and it becomes a rule the draft respects.

Your channels

Where the draft goes for approval and how each person receives their own shifts: the app you already use, text, email, Slack or Microsoft Teams.

Your systems

Which tool holds the published schedule, where availability and time off are recorded, and whether hours flow on to payroll by export or by API.

The words it uses

How a shift notification reads, how an open-shift ask reads, and the versioned prompts behind the draft — yours to read, yours to tune.

Transparent, and configurable by you

The dark band below is what the system will not do. This is what you can see and change — and it is how we build everything, not something invented for this page.

You see what it did, and why

Every assignment names the rule it satisfies, every unfilled shift names what would have to change, and the fairness report shows how the week landed on the measures you chose. You can see why somebody got Saturday before you approve it.

Versioned prompts you can tune yourself

The prompts behind the draft are yours to read and yours to tune. A change is a reviewable diff, replayed against weeks you have already scheduled by hand so you can see exactly what it would have done differently, and rolled back the same way.

A human in the loop wherever judgment actually matters

Drafting and checking are automatic. Publishing, bending a rule, approving overtime and deciding who covers a call-out are a manager's — that boundary is a property of the system, not a promise about how carefully we'll use it.

Every run logged, along with what it cost

One row per draft and one per change: the inputs it used, the rules it applied, the holes it reported, who approved it and what they edited — plus token usage and cost for that run.

It can be configured to run in any business. The coverage and the rules are yours: your stations, your rest rules, your fairness measures, your notice period. The weekly-rota wiring on this page is one configuration of it, not the limit of it.

Works with your tools

This runs on top of the scheduling and payroll tools you already use. We're not selling you a new app for your staff to install.

• 7shifts, Deputy or When I Work
• Homebase
• A time clock or timesheet tool
• ADP, Gusto or Paychex
• An HR system
• Toast or Square
• A spreadsheet you control
• Google or Outlook calendar
• SMS to each person
• Email, Slack or Microsoft Teams

Works well for

• Restaurants and cafes — Coverage changes with the weather and the bookings, and half the availability arrives as a text.
• Home care agencies — Visits have to be matched to the right qualification and the right person, week after week.
• Senior living and nursing homes — Overnight and weekend posts cannot be left short, and the rules about who may cover them are strict.
• Security guard companies — Every site has a licensed-post requirement, and a call-out has to be covered by somebody qualified.
• Retail chains — Each store builds the same grid every week against hours the district has already set.

How a rollout works

Nobody is scheduled by a bot on day one. The first milestone is a draft of a week you have already built by hand.

1. 01Discovery

   The quote call, then a working session. We go through your coverage requirement, your rules in your own words, your fairness measures, your time-off and swap policy, the notice period you have to give, and which tools hold what. You get the rule design, the exception classes and a written scope.

2. 02Dry run

   The bot drafts weeks you have already published, from the inputs you had at the time. Nothing is published and nobody is notified. We compare its draft against what your managers actually did, shift by shift, and every disagreement is either a rule we have not captured or a judgment call worth naming. This is where your own numbers come from.

3. 03Assisted mode

   Live, usually on one location or one department first. The bot drafts, the manager edits and approves, and the approved week publishes the way it does today. Building the grid stops being building and becomes checking. The control doesn't change.

4. 04Optional standing drafts

   Optional. Once a location's weeks have come through with few edits for a while, the draft can go out on a schedule for approval-by-exception, with every hard stop in force and the kill switch. Plenty of teams stop at step 03, and that's a good place to stop.

What we don't claim

• No numbers we haven't earned. No hours saved per week, no percentage fewer open shifts, no overtime-reduction figure. We haven't scheduled your team. The dry run is where those numbers come from, and they'll be yours.
• No customer claims on this page. No names, no case studies, no logos.
• It does not make your compliance somebody else's problem. The rules it follows are the rules you give it. Where the law where you operate sets rest, break, notice or minor-employment requirements, you tell us and we make them hard rules — and the manager who publishes the week is still the accountable person.
• It cannot staff a shift nobody can work. If you are short people, the draft says you are short, names who was closest and what would have to change. An optimizer that always produces a full grid is hiding something.
• Fully autonomous scheduling isn't on offer. The design assumes a manager approves every week and every change to a published week. That's deliberate, and it's why the boundary is written down.

Questions people ask

Can AI build our weekly staff schedule?

Yes - and the part that matters is what it does when the week does not fit. An agent can pull availability, approved time off and qualifications, draft every shift inside the rules you wrote, show which rule each assignment satisfies, and report how the week landed on your own fairness measures. Where a shift has nobody qualified and available, it says so and names who was closest and what would have to change, rather than producing a full grid that quietly breaks something.

Does it publish the schedule itself?

No. The draft, the unfilled shifts and the fairness report go to a manager as one screen, and they edit, approve or send it back. Nobody is texted, no app is updated and no grid is posted until that happens. The same rule holds after publication: a swap, a call-out or a late availability change produces a shortlist of who could cover it, and the manager makes the change.

Does it know the labor laws where we operate?

No, and any tool that tells you it does is selling you something. It knows the rules you give it. Rest between shifts, break requirements, overtime thresholds, the notice period you have to publish by, union terms and rules for staff under eighteen are all configuration you provide and own - we turn them into hard rules the draft cannot break, and the manager who publishes the week remains the accountable person. That is a deliberate boundary, not a gap we plan to close.

How does it handle availability and time-off requests?

It reads them wherever they actually are: the scheduling app, a sheet a manager keeps, and the texts that arrive on a Sunday night. Approved time off is a hard constraint and the draft will not schedule over it. A request that arrived after your cutoff is surfaced as a request rather than applied silently, and how many people may be off at once per role is one of your rules, not an assumption.

What happens when somebody calls out an hour before a shift?

The bot works out who could cover it - qualified, available, inside the hours, rested - in the order your rules say people should be asked, and hands a manager the shortlist with the reason each person is on it. It can send the open-shift ask in your words once the manager says go. What it does not do is reassign somebody's day on its own, because a published shift is a commitment to a person.

Will it decide who gets overtime?

No. It can flag that an assignment would cross a threshold you set, and it will keep the draft inside the maximum hours you configured, but approving overtime is a person's decision and the bot holds no part of it. The same goes for sending somebody home early, and for anything that touches pay or somebody's record.

Does it rank or score employees?

No. There is no productivity score, no reliability rating and no model opinion about who is a better employee anywhere in the design. Assignments come from availability, qualification, your rules and the fairness measures you wrote - weekend rotation, how hours are spread, whose preferred days were honored last time - and every one of those is something you can read, measure and change.

Does it work with the scheduling app and payroll we already have?

Yes, for most of them - availability and time off are read from the tool you use, a sheet you maintain, or a manager's messages, and the approved schedule publishes back into that tool, with hours flowing on to payroll the way they do today. That covers tools like 7shifts, Deputy, When I Work and Homebase, payroll systems like ADP, Gusto or Paychex, and point-of-sale systems like Toast or Square. If the schedule lives in a spreadsheet, that works the same way. Naming a system means we can read or write it, not that we are partnered with it.

Is there a record of each version of the week?

Yes. Every draft records the availability, time off, qualifications and coverage it was built from, every rule it applied, every shift it could not fill, the fairness report, the model and prompt version, who approved it and exactly what they edited - and every change to a published week is a new version with the person who made it and the reason. When somebody says they were never asked, you can check.