Staff scheduling
Making a rota with AI: what can ChatGPT do and what can scheduling software do?
Can an AI assistant such as ChatGPT make a work rota? We compare language models with scheduling software on rules, absences and time tracking.
A schedule starts with enough available hours and the right staffing. We work through a fictitious example and show what you can ask an AI assistant. We then compare building a table with managing shifts, changes and absence in scheduling software.
First check whether a schedule is possible
Making a schedule with AI starts with the input. Note availability, maximum hours, qualifications and staffing per shift. Keep preferences separate from hard requirements. Without that distinction, a proposal can look neat while being unworkable.
Fictitious example: a company has, seven days a week, a day shift of eight hours and an evening shift of seven hours. Each shift requires one senior. That requires 7 × (8 + 7) = 105 senior hours. In the input, only one employee is marked as senior, with a maximum of 38 hours and available only on working days. With this data there is no valid schedule. Senior hours and weekend availability are missing.
A useful first result is therefore a notice of missing capacity. Do not immediately ask for a completed table. First have it established whether the people listed can cover the shifts requested. Perhaps other employees are senior too, but that information must then be in the input.
A prompt that makes missing information visible
Use only this fictitious data. Do not make a schedule yet. First calculate the number of hours needed per qualification and compare this with the available hours. Name any missing availability, qualifications and rules. Do not invent employees or extra contract hours. Only make a proposal if all hard requirements can be met with the data given. Then give the hours assigned per person and the staffing per shift.
Add your table to this instruction. This is an example prompt, not a report of a ChatGPT test that was carried out. An AI assistant can also use calculation tools when working this out; therefore assess the calculation and the result rather than assuming that every chat can or cannot do arithmetic.
Check a proposal outside the chat: does every shift count, is nobody scheduled in two places at once and do the hours fit within the set limits? Also include the applicable agreements on rest, breaks and deployment. The example below is only a capacity calculation, not a check against employment law or a cao.
What scheduling software adds
Staff scheduling software can bring together availability, shifts, changes and communication. Show which data is kept up to date and which checks are really set up. Not every scheduling package generates a schedule itself, and not every automatic function uses AI.
Nostradamus describes AI support for scheduling. Use that documentation as a starting point for a demonstration with your type of staffing. Also include Atrea and Timebutler as regular scheduling or registration candidates. We are not claiming any specific AI feature, or absence of AI, for these packages.
Ask how a leave request carries through to the schedule and how an error is corrected. A nice automatic schedule is of little help if last month's availability is still being used as the starting point.
A small schedule you can check by hand
This own fictitious check example uses five working days and two shifts of four hours per day. Alex and Bo are both qualified for all shifts and may each be scheduled for a maximum of twenty hours. Both are available at all the times mentioned.
| Day | 09:00–13:00 | 13:00–17:00 |
|---|---|---|
| Monday | Alex | Bo |
| Tuesday | Alex | Bo |
| Wednesday | Alex | Bo |
| Thursday | Alex | Bo |
| Friday | Alex | Bo |
Check: ten shifts × four hours = forty staffed hours. Alex and Bo each have five shifts, so twenty hours. There are no overlapping shifts per person. This is a manually worked example, not an AI result.
Now Alex is absent on Tuesday. Bo already has twenty hours and cannot take over the extra shift within the given limit. The available capacity falls to 36 hours. Four hours of qualified cover must be added, or the staffing requirement must be changed. Have a supplier process exactly this change. Check whether the system keeps the unresolved shift visible instead of silently exceeding a limit.
Compare on the daily tasks
| Task | A standalone AI chat with your input | To check in scheduling software |
|---|---|---|
| Availability | You supply the current data | Who keeps availability up to date and when is it processed? |
| Calculating and checking | Have totals and violations shown, and check them | Which hard rules block an invalid allocation? |
| Absence | You pass on the changed situation again | Does an open shift stay visible and who can respond? |
| Publishing | A table in the chat has not yet reached your employees | How do employees receive changes and how do they confirm them? |
| Actual hours | A proposal does not record time worked | How does approved time go to time tracking or payroll? |
Start with a proposal, finish with a check
An AI assistant can help organise requirements and work out possible allocations. For repeated use, current data, traceable changes and communication are at least as important. Assess scheduling software, therefore, with a normal week and a week in which someone is absent.
Look at Atrea and Timebutler via their profiles and ask specifically about your scheduling rules. Keep the first trial fictitious. Only set a real schedule after a qualified planner has checked the staffing, availability and applicable agreements.
