Sample audit report · Cleaning company
Office and home cleaning, 26 people
This is what a company gets after the audit. The company is invented; the report was written by the AI agent. The text has not been edited.
Amounts in the text are the invented company's own figures and prices, not ours. All samples and how to request an audit
Solfjell Rent AS — what the records show
Records: 2 January – 31 March 2026.
1. In short
- Will the scheduling agent cut overtime? Not proven, and it is not the first step. The overtime comes from the way the week is built, and that can be changed for free.
- The vendor's "20% off payroll" does not hold. Overtime is 8% of what you pay the cleaners; the rest is guaranteed hours, paid whatever the schedule looks like.
- Can the planner role go part-time? No. About 129 of her 160 hours a month would remain even with the agent.
- Overtime is real: about 173 hours and 57 342 NOK a month.
- The planner is the only person who can plan. That is the risk to fix now, for 8 000 NOK.
- Lateness and "payroll faster than revenue" cannot be measured with today's records.
2. How the work runs today
Twenty-two cleaners make about 1 135 planned visits a month for 301 customers. The planner builds next week by copying last week: each customer keeps the cleaner they know, and the district comes second. By her own count the schedule takes 60 of her 160 hours a month. The other 100 go on same-day changes, complaints and cover for sick cleaners.
Every cleaner has guaranteed hours: a short day is paid in full, and anything over the day's contract hours is overtime at the rate plus 40%. Times are written on paper sheets; the app is not used, so none of the 3 514 rows of the visit list has an arrival time.
3. Where the firm loses and why
Uneven days. On average the planned day equals the contract day. But 319 cleaner-days in the quarter were planned at least an hour over the contract hours and 337 at least an hour under. The long days are paid as overtime, the short days are paid in full: about 222 guaranteed hours a month fall on days planned shorter than the contract. 60% of all overtime hours are matched by short-day hours of the same cleaner in the same week. The team lead says the same: "one day I have ten hours and the girl next to me has five".
Crossing town mid-day. 38.8% of back-to-back visits by one cleaner are in different districts. The plan leaves 60 minutes between them, against 30 inside a district. Overtime averages 0.24 hours on a day with no district change, 0.42 with one and 1.09 with two or more. That is a link in the figures, not yet a proven cause.
One person holds the schedule. In the week of 9–13 February the sheet has 1 same-day change against 7–23 in every other full week: customers' changes did not reach the cleaners. The administrator has never built a week, and keys, alarm codes, allergies and who cannot work with whom exist only in the planner's head.
Smaller things. Nobody opened the door on 112 home visits (4.7%). Free repeat visits after complaints are 105 against 3 200 visits carried out (3.3%) — not a material loss.
What the check did not confirm. That the agent cuts payroll by 20%: the figure is from a pitch slide about a much bigger company. That the planner role could go part-time.
4. What to change, and in what order
- Free, now: fixed districts for each cleaner on each weekday, and one check before the week goes out — no day far over or under the cleaner's contract hours. Start with offices: the team lead says they do not mind who comes. Some customers will get a different cleaner; that is your decision.
- 8 000 NOK, once: train the administrator as backup planner and write the routine down.
- Free: record the arrival time at every visit and note every late call.
- Free: ask the accountant for revenue and payroll by month, with payroll split into guaranteed hours, overtime and salaries.
- Not now: the scheduling agent at 6 900 NOK a month and 25 000 NOK setup. Overtime would have to fall by 12% just to cover the monthly fee, and no fall is proven. Come back to it after two months of step 1.
5. How we close this
- Uneven days and town crossings. We add two checks to your scheduling sheet: planned day against contract hours, and districts per cleaner per day. From you: the decision on fixed districts and the planner's list of customers who must keep their cleaner.
- One person holds the schedule. We sit with the planner, write her routine down and put keys, alarm codes, allergies and pairings into the sheet. From you: her time, and the decision on the backup.
- Lateness unknown. We set up a simple way to record arrival times that works without the app, and a one-line log for late calls. From you: getting the cleaners to use it.
These three setups are one package, “Order”: preparation and setup. The price is named after this review, before work starts.
- Later, if the free rule is not enough: our Workload-planning agent drafts the week and the planner approves it. It is built to order: the price is named after the review, before work starts; language-model costs are billed separately, by actual use. Trial: 4 weeks, judged on overtime hours a month.
- Nothing to sell: the decision on the vendor's agent, the accountant's figures and reminders to home customers.
- Not for an agent: complaints and late calls, which customers change cleaner, pay questions, cover for a sick cleaner on the day.
6. What this gives, and what we do not promise
- Money: no saving is promised. Overtime of 57 342 NOK a month is the ceiling, not a forecast.
- Hours: with a scheduling program the planner estimates 40 of her 60 schedule hours would go, and review adds 8.67. That is freed time on a salary, not money.
- People: no reduction. The planner role stays full-time.
7. What is missing and how to check
- How often cleaners are late and real travel time: needs arrival times.
- Whether payroll grows faster than revenue, and whether guaranteed hours were raised last year: needs the accountant's figures.
- How overtime is actually paid out, and whether three timesheet lines entered twice were paid twice: payroll should check.
- Whether customer data may go to an outside AI service, and who would look after such a tool.
The number to compare: overtime hours a month — about 173 today — with the 38.8% of visits that cross districts alongside. Look again after two months of the new rule, and decide on any agent then.