Claude can test rota and headcount scenarios from your staffing spreadsheet if you upload it with code execution and file creation turned on, state your cover rules and pay rates, and ask for each scenario as a separate costed option. Ask for an .xlsx with live formulas, then recheck the totals yourself before choosing overtime, temps or a new hire.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowThe hard part is not the prompt. It is getting the spreadsheet and the rules into a shape Claude cannot misread: one row per person, hours as numbers, holidays as words rather than colours, and every assumption written down, including what a new hire really costs before their first useful shift. Claude also cannot tell you whether a rota is legal where you trade; it can only apply the rules you give it.
The example running through this page is an illustrative gift shop with a manager and seven hourly staff facing a December staffing gap. For the different job of building a rota from hourly till sales, see how to build a staff rota from hourly till sales with AI; here the rota already exists and the question is which way to fill the gap.
What should the staffing spreadsheet contain before you upload it?
A staffing spreadsheet for Claude needs one row per person with a staff code, role, hourly rate, contracted weekly hours, maximum hours and extra availability, plus a separate tab for cover rules. Names, contact details and personal notes stay out.
Here is the gift shop's staff tab, filled in. Staff codes replace names, which keeps the upload useful while leaving out who is who:
| Staff | Role | Rate ($/hour) | Contracted hours | Max hours (owner's rule) | Extra hours available in December | Holiday weeks |
|---|---|---|---|---|---|---|
| S2 | Manager | 19.50 | 40 | 40 | 0 | none |
| S3 | Part-time | 15.00 | 24 | 40 | 8 | none |
| S4 | Part-time | 15.00 | 20 | 40 | 6 | week 3 |
| S5 | Part-time | 14.50 | 16 | 40 | 10 | none |
| S6 | Part-time | 14.50 | 16 | 40 | 4 | none |
| S7 | Weekend | 14.00 | 12 | 40 | 6 | none |
| S8 | Weekend | 14.00 | 12 | 40 | 8 | none |
| S9 | Casual | 14.00 | 0 | 40 | 10 | none |
A second tab holds the rules in plain sentences, one per line, so Claude can quote back any rule it applies. The gift shop's Rules tab for December:
Opening: Mon-Wed 9:00 am to 5:30 pm; Thu-Sat 9:00 am to 8:00 pm; Sun 10:00 am to 4:00 pm
Floor cover: Mon-Wed 2 staff; Thu-Sun 3 staff; Sat 11:00 am to 4:00 pm 4 staff
Breaks: any shift over 6 hours has a 30-minute unpaid break; someone must cover it
Manager (S2): 8 hours a week off the floor for admin
Maximum: 40 paid hours a week per person (owner's rule)
Rest: nobody closes at 8:00 pm and opens at 9:00 am the next day (owner's rule)
Pay: each person's own rate; no premiums unless listed on this tab
Writing the rules out like this takes 15 minutes and does two jobs: Claude stops guessing, and you notice rules you had only ever kept in your head, such as the close-then-open restriction.
Five preparation steps that prevent most misreadings:
- Unmerge every cell (select the range, then Unmerge Cells from the merge button's drop-down on the Home tab) and put headers on a single row. Merged headers spanning two days are the commonest cause of hours landing in the wrong column.
- Turn colours into words. A CSV is plain text, so a green fill meaning "holiday" simply disappears in that format. Add a Holiday weeks column, as above, even if you upload the .xlsx.
- Store hours as numbers, such as 8 or 8.5, not as "9-5:30" typed into a cell. Write shift times in a consistent format if you include them at all.
- Remove what the job doesn't need: names, phone numbers, addresses, dates of birth, ID numbers and any note about health or sickness reasons. Rates and hours are enough.
- Save a dated copy ("staff-model-2026-11-22.xlsx") so you can always compare Claude's output against the untouched original.
Chat upload or Claude for Excel: which suits rota scenarios?
For a one-off scenario test, upload the workbook to a Claude chat; for a model you will rerun monthly inside the same workbook, the Claude for Excel add-in is better. Both read your sheet; they differ in plan and where results land.
| Route | Plan needed | What happens | Best for |
|---|---|---|---|
| Chat with file upload | Any plan, Free included, with code execution and file creation on | Claude runs code in a sandbox on a copy of your file and can hand back a new .xlsx with working formulas | Testing three or four scenarios once, before a decision |
| A Claude project | Any plan; Free allows up to five projects | Rules and cost assumptions stay in the project, so next month's upload starts from the same ground | Rerunning the same questions each month |
| Claude for Excel add-in | Pro ($20 a month) or above | Claude edits the open workbook, keeps formula links intact and cites the cells it used | A scenario tab that lives in your real staffing workbook |
To switch on the chat route, go to Settings, Capabilities and turn on Code execution and file creation (Anthropic says the newer settings layout calls it Create and edit files). Anthropic's upload page says .xlsx uploads need this setting, and its file help page gives 30MB per file for this feature, which a staffing sheet will never approach. Expect file work to use more of your plan's usage limit than an ordinary chat.
One add-in detail matters for scenario work: Anthropic lists Excel's data tables (the What-If Analysis feature) and macros as unsupported in Claude for Excel. Build scenarios as side-by-side columns or a scenario selector cell instead. For the wider safety routine around spreadsheets, see how to use Claude with Excel and Google Sheets safely.
A gift shop's December rota, tested three ways
The shop opens 9 am to 5:30 pm Monday to Saturday and 10 am to 4 pm on Sunday. For four weeks in December it stays open until 8 pm Thursday to Saturday and needs more people on the floor. These figures are illustrative; the method is what carries over.
Step 1: let Claude count the gap (10 minutes)
The owner's cover rules for December: two staff on the floor Monday to Wednesday, three from Thursday to Sunday, and a fourth on Saturdays from 11 am to 4 pm. Claude's count of the required floor hours, checked by hand:
| Days | Working | Staff-hours a week |
|---|---|---|
| Monday to Wednesday | 8.5 open hours x 2 staff x 3 days | 51 |
| Thursday and Friday | 11 open hours x 3 staff x 2 days | 66 |
| Saturday | 11 x 3, plus a fourth person for 5 hours | 38 |
| Sunday | 6 open hours x 3 staff | 18 |
| Break cover | 14 long shifts x 30 minutes | 7 |
| Total needed | 180 |
Contracted hours add up to 140, but the manager spends 8 of those on admin, so 132 hours are available on the floor. The gap is 180 minus 132: 48 staff-hours a week for four weeks.
Step 2: cost three ways to fill it (20 minutes)
- Scenario A, extra hours only. Existing staff pick up the 48 hours from the 52 they said they could offer.
- Scenario B, two seasonal temps. Two people at $14 an hour for 24 hours a week each, hired in late November with 12 hours of paid shadowing.
- Scenario C, one permanent part-timer. One hire at $14.50 for 24 hours a week, kept after Christmas, with the other 24 hours from existing staff.
Each new hire also carries the cost-per-hire figure built in the section below: $282 before training shifts. Claude's summary, which matched the owner's own recalculation:
| Scenario | December wages for the gap | Hiring and training | December total | Spare cover a week | After December |
|---|---|---|---|---|---|
| A: extra hours | $693 a week x 4 = $2,772 | $0 | $2,772 | 4 hours | No change |
| B: two temps | $672 a week x 4 = $2,688 | 2 x $450 = $900 | $3,588 | 52 hours | Temps leave |
| C: one permanent hire | ($348 + $336) x 4 = $2,736 | $456 | $3,192 | 28 hours | $348 a week, about $1,508 a month |
Scenario A is cheapest: $816 less than B and $420 less than C. The extra money in B and C buys resilience. With only 4 spare hours a week, one person off sick for a Saturday leaves the shop short at its busiest time. Scenario B costs the most but leaves every existing person's extra availability untouched as backup.
Step 3: decide with the numbers and the parts the numbers miss
The owner chose C. January brings a sale, and two staff have holidays booked in February, so a permanent 24-hour person had work to do after Christmas. The $1,508 a month carrying cost was set against the overtime the sale would otherwise need. A different shop, with a quiet January and staff keen on extra hours, would rightly pick A and keep $420 against C. Claude's job was to make those trade-offs visible in 30 minutes instead of an evening with a calculator; the choice stayed with the owner.
If building scenario tables like these in Excel is the slow part, try The Complete Excel Course with AI Integration on Coding Liquids.
How do you write scenario prompts Claude can't misread?
Scenario prompts for Claude work when they state the gap, the rules, the pay basis and the output format, and ask for each scenario separately with its working shown. Vague prompts get averaged answers.
Start with a check of the data before any scenario. This prompt catches problems while they are cheap:
I've uploaded staff-model.xlsx with two tabs: Staff and Rules.
Before any scenarios, tell me:
1. How many staff rows, total contracted hours, and the range of hourly rates.
2. Any row with a blank or zero rate, or contracted hours above the max.
3. Any rule on the Rules tab you can't turn into a number, and why.
Don't calculate scenarios yet.
An illustrative reply, with what to do about it:
8 staff rows; 140 contracted hours a week; rates from $14.00 to $19.50. S9 has 0 contracted hours, which I've read as casual. One rule is ambiguous: "extra cover on busy Saturdays" doesn't say how many staff or which hours, so I can't cost it.
That last line is the useful one. The owner rewrote the rule as "a fourth person on December Saturdays, 11 am to 4 pm" before going further. Then the scenario prompt:
December needs 180 floor staff-hours a week for 4 weeks (rules on the Rules tab).
Available floor hours: 132. Cost three scenarios to cover the 48-hour gap:
A) existing staff only, using the "Extra hours available" column, cheapest first
B) two temps at $14/hour, 24 hours a week each, plus $450 each to hire and train
C) one permanent hire at $14.50/hour for 24 hours, plus $456 to hire and train,
with the other 24 hours from existing staff, cheapest first
Rules: pay only paid hours (30-minute breaks are unpaid); nobody over 40 hours a week;
S4 is on holiday in week 3.
Output: an .xlsx with one tab per scenario using live formulas, not typed values,
and a summary tab: December cost, spare cover hours a week, monthly cost after December.
Use each person's own rate, never an average.
Two follow-ups earn their place. A sensitivity question, "What happens to each scenario if two staff are off sick on the same December Saturday?", brought back this illustrative answer:
Losing two 8-hour Saturday shifts leaves 16 hours to cover. Scenario A has 4 spare hours in the week (S9), so Saturday runs 12 staff-hours short: two people instead of three for most of the afternoon. Scenarios B and C have 52 and 28 spare hours, but the availability column is weekly, so I can't confirm any of those hours fall on a Saturday. Add Saturday availability and I'll rerun.
The honest last sentence is the valuable part. The owner added a Saturday availability column (S7, S8 and S9 said yes), and the rerun showed B and C could absorb two absences while A could not.
The second follow-up looks past December: "With scenario C, how many hours does January's rota have spare against January's cover rules?" If the answer is close to zero, the permanent hire has real work after the peak; if it is close to 24, the new person's hours are surplus and scenario C is really a December cost with a long tail.
What should go into a cost-per-hire assumption?
A cost-per-hire assumption should include the job advert, the manager's time to screen and interview, set-up admin, uniform or equipment, and paid training shifts before the new person is useful. Leaving any out makes hiring look cheaper than overtime.
The gift shop's assumptions, filled in. Each figure is the owner's own estimate, which is exactly the point: Claude should be given these numbers, never asked to invent them.
| Item | Basis | Cost |
|---|---|---|
| Job advert | One job-board listing | $120 |
| Screening and interviews | 4 manager hours at $19.50 | $78 |
| Set-up admin (payroll, contract, till login) | 2 manager hours at $19.50 | $39 |
| Uniform and name badge | Supplier price | $45 |
| Cost per hire before training | $282 | |
| Paid shadow shifts | 12 hours at the new person's rate | $168 at $14, $174 at $14.50 |
| Cost per hire including training | $450 (temp) or $456 (permanent) |
Two refinements change decisions more than people expect. First, the manager's interview hours come out of floor time in a month when floor time is short, so a busy shop may want to count them at the overtime cost of covering the manager, not at the manager's rate. Second, the cost of a hire who leaves within a month: if one in four seasonal hires in past years has quit early, ask Claude to rerun scenario B with a 25% chance of a replacement hire and see whether it still beats C. For broader runway and hiring models, how AI can help model runway and hiring scenarios covers the longer horizon.
How do you check Claude's arithmetic before trusting a scenario?
Check Claude's arithmetic by asking for live formulas, recomputing each total yourself, confirming the hours add up to the gap, and spot-checking a few rows by hand. Ten minutes of checking protects a decision worth thousands.
- Insist on formulas. Anthropic says Claude can produce spreadsheets with working formulas. Click a total in the returned file: if the formula bar shows 2772 rather than a formula, ask again.
- Recompute the wage line. In scenario A's tab, a formula such as
=SUMPRODUCT(D2:D8,E2:E8), with hours in column D and rates in column E, should return 693 for the week. A different answer means a wrong rate or a wrong hours figure, and you can see which row. - Confirm the hours add up to the gap. Each scenario's extra hours must total 48 a week. Scenario A uses 8 + 6 + 10 + 4 + 6 + 8 + 6 = 48. If a scenario totals 46 or 50, Claude has dropped or double-counted someone.
- Check the rules held. Add a column of total weekly hours per person and use conditional formatting to colour anything above 40. Check S4's week 3 separately.
- Spot-check three rows by hand. S5 in scenario A: 10 hours at $14.50 is $145. S8: 8 hours at $14 is $112. The temp in scenario B: 24 hours at $14 is $336 a week, so two temps are $672.
A last sense check compares the answer with something you know. If the shop's December weeks normally bring in about $9,000 of sales, adding $684 a week of wages in scenario C is under 8% of sales. A scenario that claims to cost $68 a week is wrong by a decimal place, however tidy the spreadsheet looks.
Breaks, averages and colour-coded holidays: where the totals slipped
The gift shop's first run had four errors, all caught by the checks above. Each is typical of rota maths in any AI tool:
- Unpaid breaks paid. The first version priced each temp's three shifts at 8.5 hours instead of 8 paid hours, adding 1.5 hours a week per temp. Over four weeks that overstated scenario B by $168. It showed up when the temp's weekly hours came out at 25.5 instead of 24. Fix: say in the prompt that breaks are unpaid, and ask for paid hours and on-floor hours as separate columns.
- An average rate. One draft costed scenario A at 48 hours x $14.60, an average of the part-timers' rates, giving $700.80 a week instead of $693. The $7.80 looks trivial, but it hid which people were doing the hours. Fix: "use each person's own rate, never an average".
- A holiday that only existed as a colour. Before the Holiday weeks column existed, S4's week off was a green fill on the rota tab. Claude counted S4's 6 extra hours in week 3, so scenario A's 4 spare hours were really a 2-hour shortfall that week. Fix: holidays as words, in a column, every time.
- Four-week months. Scenario C's ongoing cost came back as $1,392 a month (4 x $348). A year has 52 weeks, so an average month is about 4.33 weeks and the real figure is about $1,508: $116 a month, or $1,392 a year, missing from the decision. Fix: ask for annual cost divided by 12.
Where does Claude's job end and employment advice begin?
Claude's job ends at arithmetic and applying the rules you supply; whether those rules match employment law, contracts and pay requirements where you trade is a question for an employment adviser or official guidance.
In one run, Claude added an unprompted line about whether seasonal staff earn paid holiday. Whatever such a line says, it is not a modelling input. Strip it out, and take questions like these to someone qualified before you act on a scenario:
- Can staff agree to work above their contracted hours, and is a premium rate due in their contracts or under the rules that apply to you?
- What rest is required between a late close at 8 pm and a 9 am open the next day?
- What must a fixed-term seasonal contract say about notice, holiday and end date?
- Do any staff have pay rates set by age or by a collective agreement?
- Can you change a permanent person's contracted hours after the peak, and with how much notice?
Once you have the answers, write them on the Rules tab as constraints ("no close-then-open shifts for the same person", "S3 and S4 premium of $2 an hour above 37.5 hours") and rerun the scenarios. That keeps Claude doing what it is good at. The companion tutorial on rules to enter so an AI-built rota stays legal covers how to phrase them.
Can you reuse the model for January and next year's headcount?
A rota scenario model can be reused if the rules and cost assumptions live in one place, such as a Claude project or a Rules tab, and each month you only swap in new hours, availability and holidays.
A sensible monthly routine for the gift shop, about 20 minutes once the model exists:
- Update the staff tab: rates, contracted hours, availability, holidays.
- Update the cover rules if opening hours change (sales, late nights, public holidays your shop observes).
- Upload to the same project and run the saved scenario prompt.
- Run the five checks above; most months only the hours total and one spot-check are needed.
- Save the returned workbook with the month in the file name, next to the decision you made.
For headcount over a year, ask a different question: "Using the last 12 months of weekly sales on the Sales tab and the rule of 20 floor staff-hours for every $1,000 of weekly sales, never fewer than two on the floor, how many contracted hours do we need each quarter, and where do we run short?" At December's $9,000 a week, that rule gives the same 180 hours counted above. That turns a series of December panics into one planning conversation in September. AI shift scheduling versus a rota spreadsheet is worth reading if the monthly routine starts to feel like the job a rota app should do.
If you want help fitting staffing scenarios into how your shop runs week to week, I offer one-to-one AI consultations for small businesses.
What does it cost to run rota scenarios in Claude?
Running rota scenarios in Claude can cost nothing on the free plan, which includes file analysis; Pro at $20 a month ($17 a month billed annually) adds more usage and the Excel add-in.
| Plan | Price | Fit for this job |
|---|---|---|
| Free | $0 | Fine for a one-off test of three scenarios; usage limits may cut a long session short |
| Pro | $20 a month, or $17 billed annually | Comfortable for monthly reruns; adds Claude for Excel |
| Team | $25 a seat a month, or $20 billed annually, minimum two seats | When the manager also runs the model; no training on your content by default |
The data question matters more than the price for staff records. On individual plans (Free, Pro, Max), Anthropic uses chats to improve its models only if you allow it, through the Model Improvement setting in your privacy settings; Team and Enterprise are listed as not training on your content by default. With staff codes in place of names and no personal details in the sheet, the upload carries rates and hours, which is all the scenarios need.
Questions about rota and headcount scenarios in Claude
Can the free Claude plan analyse a staffing spreadsheet?
Yes. Anthropic says code execution and file creation is available on every plan, Free included, and it is the feature that analyses uploaded spreadsheets. Turn it on under Settings, Capabilities first; Anthropic's upload page says .xlsx uploads need it. The Claude for Excel add-in, which works inside the workbook, needs a paid plan from Pro at $20 a month.
Should I upload my rota as CSV or as an Excel file?
Upload the .xlsx if your workbook has several tabs, such as staff, rota and rules, because a CSV holds one sheet only. A CSV is fine for a single flat table. Either way, colours do not survive as data: a holiday marked by a green fill must become a word in a column before Claude can count it reliably.
Can Claude tell me whether my rota breaks employment rules?
No, not reliably. Claude can apply rules you write down, such as a 40-hour weekly cap or a minimum gap between a late close and an early open, and flag shifts that break them. Whether those are the right rules for your staff, contracts and location is a question for an employment adviser or official guidance.
How accurate are Claude's rota cost totals?
Accurate when the inputs and rules are clear, but always check them. Typical slips are paying unpaid breaks, using an average hourly rate instead of each person's rate, and converting weekly costs to monthly with four weeks instead of about 4.33. Ask for an .xlsx with live formulas, then recompute the totals yourself with SUMPRODUCT.
Can Claude build the rota itself, shift by shift?
Claude can draft a shift-by-shift rota from your cover rules and staff availability, and it is a reasonable first draft for a small team. Treat it as a draft: check each person's hours, rest gaps and holidays, then publish it through your normal rota tool so staff see one version. Scenario testing, not shift allocation, is where Claude saves most time.
Further reads
- Restaurant Labour Cost Percentage: How to Test a Rota Change — Test a rota change against labour cost as a share of sales.
- Can AI Automate Rotas, Timesheets and Leave Requests? — What AI can automate in rotas, timesheets and leave requests.
- Claude for Excel Cost: Which Claude Plan Do You Need? — Which Claude plan you need for the Excel add-in route.
- Staff Turnover Analysis With AI: Keep Employee Records Private — Analyse staff turnover in a spreadsheet while keeping records private.
- How to Clean Messy Data in Excel: Spaces, Duplicates, Text — Tidy a messy staff sheet before any AI reads it.
- How a Staffing Agency Can Use AI to Forecast Client Demand — Turn two years of placement history into a seasonal demand forecast and a bench target, using a spreadsheet and AI for the clean-up and cross-check.
- Nursery Staff Rota: How to Build One That Keeps Ratios With AI — A nursery rota built from the EYFS ratios: a room-by-room grid, break cover, absence swaps and the checks a manager still does by hand.
Browse all AI-for-business tutorials
Sources: Anthropic help centre articles on creating and editing files and on uploading files; Anthropic's Claude for Excel guide; claude.com/pricing; Anthropic's privacy centre article on model training. All read 22 November 2026. Business figures are illustrative.
