For a restaurant, the features that matter are split-shift handling, tip-pool rules, shift swaps with manager approval, availability, a labour-cost view against forecast sales, and break and rest-period flags. The AI forecast is helpful but secondary. Test each with a trial week on your own rules before paying.
Generic scheduling lists rank for this search, and they treat a restaurant like any shift business. Yours differs in four ways: staff work lunch and dinner with a gap in between, tips are shared by formula, swaps happen daily, and sales swing with weather and bookings. An app earns its fee by handling those without a spreadsheet beside it.
Run one trial week before you compare anything
Feature pages are written for the demo. To see how an app behaves, give every candidate the same awkward week. Prepare it once, in a plain document, and run it in each free trial.
The trial pack needs six things: your 12-15 real staff roles and hourly rates (use made-up names or initials), their availability including two people who cannot work Sunday lunch, one split-shift pattern, your tip-pool rules in one paragraph, a forecast for each shift on the busiest and quietest day, and your break rules as the manager understands them. If the vendor will not let you enter these during a trial, that tells you something.
Then score each app on the sheet at the end of this tutorial. A good result is not a perfect rota. It is an app where every failure is visible, fixable by a manager in a minute, and explained on the screen.
Split shifts: three tests that catch the weak apps
A split shift is two blocks in one day with an unpaid gap, such as 11:00 to 14:30 and 17:00 to 22:00. The rota looks simple. The trouble comes in what the app does with it.
- Does it count as one shift or two? The two blocks above total 8.5 paid hours. Check the app shows 8.5 hours for that person for the day, not two separate entries that each look under any overtime threshold. Overtime and premium-pay rules often depend on the day total or the span, and the rules vary where you trade, so confirm yours with a qualified adviser.
- Does it flag short turnarounds? A member of staff who closes at 00:30 and opens at 06:30 has six hours between shifts. Many restaurants set their own minimum rest and some places legislate one. Enter one of these on purpose and see whether the app warns, blocks, or says nothing.
- Can staff state split-shift availability? Some people will work lunches only; others want a long gap for a second job. An app that records only "available 10:00-23:00" will keep offering that person a split they refuse.
A realistic failure: the manager drafts Friday, the AI places a server on both a 10:30 lunch and a 17:00 dinner, and the app's weekly view shows two tidy blocks. The server, who is a student, comes to the manager on Thursday to say the 2.5-hour gap is no use to them. A good app would have known that from their availability and never suggested it.
Tip pooling: test it with numbers where the pennies don't divide
Tip rules differ by place and by restaurant. Some share by hours worked, some by points per role, some pool front of house only, some include the kitchen where the law allows. Whatever your rule, what you need from an app is that it applies your formula, shows its working and sends a clean figure to payroll. Ask the vendor to describe the formulas it supports before the trial, and enter a pool with awkward numbers.
An illustrative test. A Saturday pool is $1,000, and your rule gives each role a weight per hour worked: server 1.0, food runner 0.6, busser 0.5.
| Person | Role | Hours | Weight | Weighted hours | Share of $1,000 |
|---|---|---|---|---|---|
| A | Server | 30 | 1.0 | 30.0 | $402.68 |
| B | Server | 25 | 1.0 | 25.0 | $335.57 |
| C | Runner | 20 | 0.6 | 12.0 | $161.07 |
| D | Busser | 15 | 0.5 | 7.5 | $100.67 |
The total weighted hours are 74.5, so each weighted hour is worth $13.4228. The rounded shares sum to $999.99, one cent short. Run the same figures in the app and look for two things: does it show the per-hour rate, and does it handle the missing cent (by rounding to the largest share, or carrying it to the next pool) rather than quietly losing it? A one-cent difference looks trivial, but a payroll export that does not reconcile to the pool is the sort of defect that turns into an argument with a team member.
Three more tip checks. First, change a role mid-shift (a server who covers the bar for two hours) and see if the app can split the hours between weights. Second, remove a person who left early and see whether the pool recalculates. Third, look at who can see the figures. Staff should see their own tips; managers should see the pool; neither should have to export to a spreadsheet and email it.
Swaps and cover: the approval trail matters more than the button
Almost every app lets staff post a shift for someone to pick up. What separates them is the trail and the guardrails. Run these scenarios during the trial:
- A clean swap. Two servers of equal skill swap Tuesday dinner. The app should ask for manager approval, or let you set auto-approval for same-role swaps, and update the rota and the labour cost.
- A swap that creates overtime. The server picking up a shift is already on 38 hours. The app should warn the approver before the approval, not after payroll.
- A swap that breaks the skills mix. A new starter takes the senior bartender's Saturday. A good app lets you restrict pick-ups by role or skill.
- The ignored request. A shift is posted, nobody takes it, and it falls to the manager at 4pm. Does the app nudge the people who might say yes, and does the manager see an open-shift list?
Toast IQ Team's own page, which I read on 27 October 2026, says staff use the MyToast app to view schedules, swap shifts and check earnings. That is a useful example of the category, and a reminder to check whether a swap can be limited by role and whether approvals are logged. It is a description from the vendor's page, not a test result.
What the AI part should actually do
Scheduling apps now use "AI" for several different things. Separate them before you pay for one.
- Demand forecast. Predicts covers or sales per shift from past POS data, bookings and sometimes weather or events. It only helps if the data is connected. Toast's page describes forecasting from historical POS data, and a scheduling agent that drafts a schedule from availability and projected sales, which the manager reviews and publishes. Ask any vendor what history it needs before it is useful; a restaurant open for six weeks has little to learn from.
- Auto-draft. Builds a first rota from availability, hours caps and the forecast. Judge it by how many edits you make to the draft. If you change a third of the shifts, the time saved is small.
- Assistant questions. A chat box that answers "what drove overtime last week?". Toast says its IQ assistant answers labour questions such as overtime and the effect of weather or events. Check whether answers cite the underlying numbers you can see elsewhere in the app.
- Compliance flags. Break and rest warnings. Often rule-based rather than AI, which is fine. What matters is whether you can edit the rules for how you operate.
If you only want to try the forecast idea without buying an app, the tutorial on building a restaurant rota with AI demand forecasts shows a do-it-yourself version, and forecasting covers to cut food waste reuses the same numbers for ordering.
The labour-cost view: test it against your own target
A restaurant manager makes the scheduling decision with a percentage in mind: labour cost as a share of sales. The app should show that number live as you move shifts, using forecast sales rather than last week's actuals.
An illustrative Friday. Forecast sales are $4,200 and the target is 28%, so the labour budget is $1,176. The drafted rota has three servers at six hours and $12 an hour ($216), two cooks at eight hours and $18 ($288), one dishwasher for seven hours at $14 ($98), a runner for five hours at $12 ($60), and a manager for eight hours at $25 ($200). Wages total $862. Add 12% for employer costs and it is $965.44, which is 23.0% of forecast sales. Comfortable.
Now the forecast is cut 15% because of weather, to $3,570. The same rota is 27.0%. Two questions for the trial: does the app recalculate when you change the forecast, and does it let you add on-costs so the figure matches your payroll rather than only base wages? An app showing 24.1% (base wages over the reduced forecast) when your real figure is 27.0% will flatter every rota and mislead the team.
Also check whether the view splits by day part. Lunch can run at 35% and dinner at 22% while the day looks fine, and that split tells you where to trim.
Availability, time off and the Sunday problem
Availability looks like a basic feature until you need it. Test these: a recurring weekly availability, a one-off change, time-off requests with approval, and what the draft does when a request arrives after the rota is published. Also test the awkward case: two people want the same Saturday off and you can spare only one. The app should show who is already off, not leave you to remember.
Ask how staff report availability. If they have to log in to a web page, uptake will be low. A phone app with push notifications tends to be used; an email-only flow is the slowest option. Ask a few team members to try the login during your trial, because the person who will not use the app is a better test than the manager who will.
Break and rest rules: make the app prove it
Break, meal and rest rules vary widely, and an app that says it is "compliant" cannot know where or how you operate. Treat these as settings you configure and verify, then have a qualified adviser look at the configuration if you have any doubt.
- Can you set a paid or unpaid break by shift length (say, 30 minutes unpaid after six hours)?
- Does the app warn when a drafted shift has no break?
- Does it record when the break was actually taken, or only scheduled?
- Can minors, if you employ any, have different limits?
- Does the clock-in sync from the till, so a manager sees early clock-ins and late finishes against the rota? Toast describes this sync for its own system; ask any independent app whether it connects to your POS and what that connection costs.
Scoring sheet: one row per feature, one column per app
Weight the rows by what hurts if it fails. An illustrative filled-in version for a 14-person restaurant, scoring 0 (fails), 1 (works with workarounds) or 2 (works cleanly):
| Feature | Weight | App 1 | App 2 | POS tool |
|---|---|---|---|---|
| Split shifts counted as one day | 3 | 2 | 1 | 2 |
| Tip pool formula and rounding | 3 | 1 | 2 | 2 |
| Swaps with approval trail | 2 | 2 | 2 | 2 |
| Labour % with on-costs | 3 | 0 | 2 | 1 |
| Forecast uses your sales history | 2 | 1 | 1 | 2 |
| Break and rest warnings | 2 | 1 | 2 | 1 |
| Staff phone app is easy | 2 | 2 | 1 | 2 |
Weighted totals: App 1 is 6+3+4+0+2+2+4 = 21; App 2 is 3+6+4+6+2+4+2 = 27; the POS tool is 6+6+4+3+4+2+4 = 29 out of 34. The POS tool leads on this sheet, mainly through tip pooling and the forecast, while App 2 beats App 1 on the cost view and breaks. The totals are a starting point, not a verdict: if App 2 costs half as much, a four-point gap may not justify the POS tool, and a zero on a heavily weighted row, such as App 1's labour view, can disqualify an app whatever else it scores. Put the quotes beside the scores, then let contract length and exit terms decide between the close ones. The questions in twelve questions to ask a restaurant AI vendor before signing help with that.
Run the sheet twice if you can: once straight after the trial, and again after two weeks of real use. Scores change when the novelty wears off. An app that scored 2 for the staff phone experience on day one may drop to 1 when half the team has not logged in by day twelve, and that second score is the one to trust.
Price, contract and what to ask about the POS tie-in
Scheduling apps usually price per location or per employee per month, with tiers for payroll, tips or labour reports. I have not stated a figure for any here because plans change often and several vendors publish prices only after a sales call; for example, Toast's IQ Team page lists no price and links to a separate pricing page. Put the quotes in your scoring sheet in USD per month for your headcount, and ask three questions of each: is the price per active employee or per seat, are payroll and tips separate add-ons, and what happens to the data if you leave.
If you already use a POS, ask whether its scheduling is included or an extra, and whether an independent app can read your sales and clock-in data without a third-party connector fee. The tutorial comparing POS-built AI with add-on tools goes into that decision, and free scheduling plans are worth a look if your team is under about ten people. For a head-to-head on cost alone, see six rota tools compared by cost.
Failures that show up in week two
- The forecast trusted blindly. A private party books 40 covers on Saturday. The model has never seen it, so the draft staffs a normal night. Add known events by hand and check the app lets you.
- Everyone is "available". Staff leave default availability in place, and the AI fills the rota from data that was never true. Ask for a one-off availability refresh at the start.
- Tips paid late because the export did not reconcile. The cent problem again, or a role mapping that nobody set. Reconcile the first pool to a calculator before you trust the second.
- Swaps agreed in the group chat. If staff still arrange cover on messaging apps, the rota is wrong by Monday. Make the app the only place a swap counts.
- Payroll not matching. The app's totals differ from the payroll run. Find out in week one, with a single pay period's data, not in month three.
When you have picked, roll out slowly. Publish the first live rota from the app with the old method running in parallel for one cycle, and ask the team to report every mismatch. If you would rather have someone plan the trial and the scoring with you, that is the sort of job an AI implementation consultation covers.
Questions before you pick a scheduling app
Does an AI scheduler replace the manager's judgement?
No. Tools that draft a rota from availability and forecast sales give you a first version. A manager still has to review it for skills mix, a new starter who needs a senior beside them, and the private-party booking the forecast cannot see.
Is the POS vendor's own scheduling tool the safer choice?
It has the advantage of live sales data and clock-ins synced from the till. It is not automatically better at your tip rules or swap approvals. Test it on the same trial week as any independent app before deciding.
How long should a trial run before paying?
At least one full week of real rota building, including one swap, one late availability change and one tip pool run on dummy numbers. A demo call on the vendor's prepared data will not show you the awkward cases.
Should staff see each other's availability?
Usually no. Staff should see their own availability, their rota and open swaps, while managers see everything. Check the permission settings during the trial, because defaults differ between apps.
Further reads
- Restaurant Labour Cost Percentage: Test Rota Changes With AI First — Testing rota changes against your labour-cost target first.
- Restaurant Tip Pooling: Can AI Draft a Fair Policy Under the Rules? — Drafting the tip-pool policy the app will apply.
- AI Shift Scheduling vs a Rota Spreadsheet for Small Teams — When a rota spreadsheet is still enough.
- Hotel Staff Scheduling With AI: Front Desk, Housekeeping, Breakfast — A department-by-department rota method: occupancy sets the cover, hard rules limit the AI, and a checker sheet catches what its draft gets wrong.
- Can AI Draft a Restaurant Staff Handbook? What an Adviser Must Check — Sort each handbook section into AI-safe, owner-facts or adviser-only, then see the tips, leave, shift and version-control checks to finish before issuing.
- Restaurant Job Descriptions: Can AI Write a Server or Line Cook Ad? — ChatGPT can draft a server or line cook ad in a minute. Here is what a hiring owner must check on shifts, pay, tips and wording before posting.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Toast IQ Team and Toast IQ product pages (checked 27 October 2026); each scheduling vendor's own help pages, which you should check during your trial.