Run a two-week test first: put every repair request into one shared form and sheet, let an AI sort each by urgency against written rules and draft a weekly chase list. If requests are getting lost, a free form and sheet fix it. If you need photos, scheduled checks and several technicians, software is worth paying for.
The test tells you which problem you actually have. Most late repairs have one of three causes: the request was forgotten, nobody owned it, or it is waiting on a part or a contractor. A sheet and a chase list fix the first two. Neither a sheet nor software fixes the third, which is why the test comes before the purchase.
Four ways repairs go wrong, and which tool cures which
A dripping tap becomes a lost booking in several different ways, and each needs a different cure. Sorting your own repairs into these four types is the real job of the test.
| What goes wrong | How it looks | What cures it |
|---|---|---|
| Forgotten | A guest mentions a fault at the desk, someone says they will tell maintenance, and nobody does | One form, one sheet and a weekly chase list |
| Unowned | Everyone knows about it, so nobody is named; two people each assume the other has it | A rule that every request gets an owner and a target date the day it is logged |
| Waiting | The fault is known and owned, but needs a part, a contractor or an approval | Neither tool: spare parts, supplier terms and phone calls |
| Repeated | The same room reports the same fault three times in a month | A repeat flag, then a proper fix of the cause |
Software bought to cure the third problem is the expensive mistake, because the fault list looks tidier while the rooms stay broken. The two-week count separates the types, so you can see which one is costing you nights.
A two-week test, set up in one afternoon
The only rule that matters is one front door. Every request, from any member of staff, goes through the form. A guest who tells the desk about a fault gets it typed in by the desk. A message sent to the maintenance person's phone is not a request until it is logged; for the first week, the maintenance lead can log those on the sender's behalf, but nothing is exempt. Set aside 60 to 90 minutes with a free forms tool and a spreadsheet.
| Form field | Type | Why it is there |
|---|---|---|
| Where | Room number or area (lobby, corridor, restaurant, outside) | Groups repeats by place |
| What kind | Drop-down: plumbing and hot water, electrical and lighting, TV and Wi-Fi, heating and cooling, locks, doors and fittings, decor and damp, lifts and exterior, other | Lets you count by type later |
| What you saw | One short sentence | Gives the assistant something to read |
| Is a guest affected right now? | Yes or no | Drives urgency |
| Can the room be sold tonight? | Yes, no or not sure | Starts the out-of-order record |
| Photo | Optional link or attachment | Saves a trip to look |
The sheet behind the form gains columns your team fills in: owner, urgency (the assistant's suggestion and your final answer), target date, status, date closed, and a blocker column with the choices "none", "part" and "contractor". That blocker column is the most valuable of them, because it tells you which of the four problems each late request belongs to.
Urgency rules the AI applies, and the ones it gets wrong
Write four levels down in plain words before anyone sorts anything, because an assistant can only apply rules it has been given. These are illustrative and need your own targets:
- P1, same day: any safety risk (a smell of gas or burning, smoke, sparks or exposed wiring, a flood, a fire door or alarm fault, a lock or window that will not secure), or no hot water (lukewarm counts), no heating in cold weather or no cooling in hot weather in an occupied room.
- P2, within 24 hours: a room that cannot be sold tonight, a fault affecting more than one room, or any fault a guest has already complained about.
- P3, within three days: a fault in a room that can still be sold, such as a TV, a kettle, a slow drain, a loose handle or a flickering light.
- P4, within 14 days: cosmetic or out of guest sight, such as paint marks, tired grout or a squeaky door.
You classify repair requests for a 24-room hotel. Use these rules.
P1 (same day): any safety risk (gas or burning smell, smoke, sparks or exposed
wiring, flooding, a fire door or alarm fault, a lock or window that will not
secure), or no hot water (lukewarm counts), heating in cold weather or cooling in
hot weather in an occupied room.
P2 (24 hours): a room that cannot be sold tonight; a fault affecting more than one
room; any fault a guest has already complained about.
P3 (3 days): a fault in a room that can still be sold.
P4 (14 days): cosmetic or out of guest sight.
If the wording is vague or casual, choose the HIGHER urgency and say why.
If a request mentions two faults, return two rows.
If it looks like a repeat (same place and kind in the last 14 days), give the
earlier ID.
Return: ID | urgency | one-line reason | repeat of ID or "no" | room out of order
tonight? (yes, no or ask)
REQUESTS:
[paste the new rows]A small sample of the answers, with my correction beside each (illustrative):
| Request as logged | AI said | Final | Why I changed it |
|---|---|---|---|
| Smell of gas by the kitchen door, guests walking past | P1 | P1, act first | Correct, but this is not a ticket. Stop, ventilate and call the emergency number, then log it. |
| Room 112 shower only lukewarm, guest in until Friday | P3 | P1 | "Only lukewarm" sounded mild, but the rules count lukewarm as no hot water in an occupied room. |
| Fire door on the second floor not closing fully | P2 | P1 | A safety fault, whatever the wording. |
| Room 210 key card reader works on and off, guest locked out twice | P2 | P1 | Security and a guest affected. |
| Room 305 TV remote has no batteries | P3 | Housekeeping, no ticket | Not a repair at all; send it to housekeeping. |
| Lobby paint scuff | P4 | P4 | Correct. |
Four of six needed changing. Three were rated too low because casual wording hid a P1 fault: weak hot water in an occupied room, a fire door and a failing lock. The fourth was not a repair at all. The fix for under-rating is a keyword override, written into the rules and not left to judgement. Anything that mentions a smell, gas, burning, smoke, sparks, flooding, a fire door, a fire alarm, exposed wiring, a lock or key card that fails, or lukewarm water in an occupied room is P1 until a person says otherwise, whatever else the wording says. The assistant's job is to sort the ordinary requests quickly. A person confirms every P1 and P2, and only the general manager or maintenance lead can downgrade one.
For the same triage idea applied to rented homes, see how property managers use AI to triage maintenance requests; the hotel version adds one more question, whether the room can be sold tonight.
Rooms out of order: count the nights, not the rooms
In many property systems an out-of-order room is removed from sale and from the room count, while an out-of-service room stays in the count and is expected back soon. Vendors differ on how each status feeds reports such as occupancy and RevPAR, so check how yours treats them, and make the sheet match the system. What the sheet adds is the number the system rarely shows: how many nights each fault cost, and how many of those were nights you could have sold.
Keep a second tab with one row per blocked room: the room, the date it left sale, the date it returned, the request ID, the cause, the nights blocked and, for each night, whether the hotel was full or turning guests away. Here is the fortnight from a 24-room hotel with an average rate of $128 (illustrative):
| Room | Nights out | Cause | Sold-out nights among them | Saved by a faster chase |
|---|---|---|---|---|
| 204 | 3 | No hot water; waiting for a part | 1 | 0 |
| 112 | 2 | Air-conditioning fault, repaired by the contractor on schedule | 0 | 0 |
| 107 | 4 | Ceiling leak; the plumber was not chased until day three | 2 | 2 |
| 209 | 2 | Door lock; nobody owned the request | 1 | 1 |
| Total | 11 | 4 | 3 |
The sums are short. Four sold-out nights with a room missing, at $128 each, is $512 of room revenue at stake in the fortnight, and three of those nights, $384, were avoidable because a request had been forgotten or left without an owner. Only nights when the hotel was full or turned someone away count as lost; a blocked room on a quiet Tuesday costs nothing in revenue, though it still costs a guest's goodwill if it was sold and then moved. The table also ranks repairs honestly: a leak that keeps a room out for four nights, two of them sold out, deserves a chase before a stained carpet does.
The Monday chase list
Once a week, give the assistant the log and ask for a list organised by who has to act. It sorts and counts; a person sends the messages.
Here is the repair log as a table (columns: ID, date, place, request, urgency,
owner, target date, status, blocker). Today is Monday [date].
Write this week's chase list:
1. OVERDUE: every open item past its target date, grouped by owner, oldest first,
with the days overdue.
2. DUE THIS WEEK, grouped by owner.
3. WAITING: items with a blocker (part or contractor), with how long each has
waited and the one question to ask.
4. ROOMS OUT OF ORDER: each room, nights so far and the target return date.
5. REPEATS: places with two or more requests of the same kind in 30 days.
Use only what is in the table. Do not invent dates, names or causes. End with three
one-line messages: one to the maintenance lead, one to a contractor, one to the
general manager.A section of the output for the test hotel looked like this (illustrative):
Overdue, maintenance lead: #12, basin tap in 204 (P2), 4 days overdue; #19, wardrobe door in 108 (P3), 2 days overdue. Waiting, contractor: #23, boiler pressure, waiting 3 days; ask for a visit date today. Rooms out of order: 107, 2 nights so far, target return Wednesday. Repeats: room 210 has had two lock faults in 30 days.
Message to the contractor: Hello, the boiler pressure fault logged on Thursday is still open. Could you confirm a visit date today? It affects the whole second floor.
The correction was in the last sentence. "It affects the whole second floor" was not in the log, which only said hot water was slow in two rooms. The instruction not to invent causes was in the prompt, and one slipped through anyway, so read each message before it leaves. A good chase message names the room, the fault, the date it was first reported and the date you need it fixed, and nothing more.
What the test showed in a 24-room hotel
After 14 days the sheet held 47 requests, about 3.4 a day. The urgency split, after a person had confirmed each, was 4 at P1, 11 at P2, 19 at P3 and 13 at P4. Here is where they stood on day 14 (illustrative):
| Status at day 14 | Requests |
|---|---|
| Closed within the target time | 31 (66%) |
| Closed, but late | 5 |
| Open and already overdue | 8 |
| Open, not yet due | 3 |
| Total | 47 |
The 13 late requests (5 closed late plus 8 overdue) are the interesting ones, and the blocker column sorts them:
| Why it ran late | Requests | Which tool helps |
|---|---|---|
| Forgotten, or no owner named | 6 | A form, an owner rule and the Monday list |
| Waiting for a part | 3 | Neither: hold spares, change supplier |
| Waiting for a contractor | 3 | Neither: agree response times, chase by phone |
| Duplicate or unclear | 1 | A repeat check and a clearer form |
Six of the 13 are exactly what a free form and a weekly list should cure, and six would have waited whichever tool the hotel owned. That reading shifts the question. The hotel's real choice was between a sheet it could start on Monday and a paid tool that would leave half its delays untouched. It ran the sheet for a month before deciding anything else.
What software adds that a sheet cannot, and what it costs
None of this means software is a waste. It does several things a form and a sheet do badly:
| Need | Form, sheet and AI | Maintenance software |
|---|---|---|
| Photos from a phone | An optional attachment; clumsy | Built in |
| Alerts to the technician | You send a message | Automatic notifications |
| Overdue reminders | Your Monday list | Rules that nudge or escalate |
| Scheduled checks and history per asset (boiler services, filter changes) | A calendar and checklists | Recurring tasks with a record for each asset |
| Parts, purchase orders, time and cost per job | More sheets | Modules, usually on higher plans |
| Room status updating the property system | Typed in twice | Only where an integration exists |
As of October 2026, the general maintenance tools publish per-user prices. MaintainX lists a free Basic plan with unlimited requesters, Essential at $20 per user a month billed annually ($25 monthly) and Premium at $65 ($75 monthly), where parts inventory, purchase orders and time and cost tracking appear. UpKeep shows Essential at $24 and Premium at $55 per user a month (the page does not say whether that assumes yearly billing), with free requester accounts for people who only submit and track requests. Both pages show a dollar sign without naming the currency, so confirm at sign-up.
Hotel-specific tools sell differently. Flexkeeping, for example, describes housekeeping, maintenance requests, tasks and guest requests in one app and lists integrations with several property systems, including Mews, Cloudbeds, Apaleo and Oracle's Opera Cloud, but it publishes no prices and invites a demo. Ask any such vendor for a written quote and for what each answer would change: who counts as a user (are desk staff who only report faults free?), whether the property-system link works both ways, what setup costs, how long the contract is and how you export your data if you leave. The pre-switch checklist for hotel management software covers the wider questions.
A quick sum puts the fee in proportion. Three users on MaintainX Essential billed annually cost $20 a month each, so $720 a year, and three on UpKeep Essential at $24 a month each cost $864. In the test hotel, one fortnight's avoidable out-of-order loss was $384. That tells you the problem is worth solving, not that software solves it, because the same avoidable nights came from forgotten requests, which the chase list attacks for free. Run the list for a month. If avoidable nights keep appearing anyway, the case for paying is strong.
A decision table built from your own test numbers
| What the test showed | What it means | Best move |
|---|---|---|
| Most late requests were forgotten or unowned | A process gap | One form, an owner on every row and the Monday list. No software yet. |
| Most late requests were waiting for parts or contractors | A supply gap | Hold fast-moving spares, agree response times in writing and phone rather than email. Software will not help. |
| Around 40 or more requests a week, or three or more people sharing repairs | A coordination load | Trial software with free requester accounts, and keep the form until it proves itself. |
| Staff will not use the form on their phones | An adoption gap | Shorten the form, put a QR code in the staff room, or give two staff a better phone app for two weeks. |
| You must keep recurring checks and records, such as service dates and inspection logs | A records need | Software with recurring tasks and an audit trail, or a shared calendar with checklists if volumes are low. |
| Rooms out of order for many nights | A revenue gap | Rank repairs by sold-out nights at stake, not by who asked, and chase those first. |
The 40-a-week line is a working rule for a small hotel, so set your own from the test. More important is to fix the trigger for buying before the test ends, so the numbers decide and the loudest voice does not. For example: "Buy if, after a month of using the chase list properly, more than a quarter of requests are still late for reasons that are not parts or contractors, or if we need recurring checks with records."
Five slips from the test fortnight
These come from the illustrative fortnight above, and they are the kind that turn up in most hotels.
- A casual description of a serious fault. "A funny smell by the back stairs" was rated P3. The keyword override now catches smell, gas, burning and smoke, and the maintenance lead looks at any P1 within the hour.
- Everything marked urgent. The first version of the form had an "urgent?" tick-box, and in week one staff ticked it on almost two thirds of requests, which made it meaningless. The box is gone: staff now describe the problem and answer "is a guest affected?". The assistant suggests the level and a manager confirms the top two.
- Duplicates. The same dripping tap was reported by three people on three days. The repeat check links new rows to the first ID and adds the reporters to it.
- The form nobody used. By day four, requests were again arriving by message to the maintenance lead. The answer was the front-door rule: "please log it", and for the first week the lead logged them on the sender's behalf and said so.
- A blocked room on the cleaning plan. A room out of order for a leak still appeared on the day's list, and an attendant spent 30 minutes cleaning it. A daily subtraction of blocked rooms from the list fixes that; allocating rooms by arrival time shows where that list is built.
Open faults also belong in the shift handover, so the evening team knows which rooms are blocked and why. Turning desk logs into handover notes with ChatGPT handles that side, and AI for hotel housekeeping checks and maintenance requests puts this job next to the others.
What stays with a qualified person
The assistant sorts requests. It does not decide whether anything is safe or compliant. Safety checks that your rules require, such as fire alarm tests, water safety checks and gas or electrical certificates, are set by the rules where you operate and done by competent people. Keep the official records for those separate from the request log, in whatever form the rules require, and use a qualified contractor for gas, electrics and structural work. If a request describes something that could hurt a person, the right first move is a phone call to someone who can fix or make it safe, and the right second move is the log entry. The form is there to make sure nothing is forgotten, not to slow down an emergency.
Questions about hotel maintenance software and AI
Is a spreadsheet enough for a 20-room hotel?
Often, yes, if one or two people do the repairs and every request goes through one form. The test in this tutorial answers it for your hotel. If requests still get lost or duplicated after two weeks of honest use, or you need photos, scheduled checks and several technicians, that is the point to trial software.
Should guests be able to report faults through the form?
Not during the test. A guest-facing QR code brings in requests worded for comfort rather than repair, and mixes service asks such as towels or pillows with real faults. Let the desk log what guests report, so staff can check the room and reword the request, then decide after the fortnight whether a guest form would add anything the desk does not already catch.
Can AI chase contractors for me?
It can draft the message and list what is overdue, but a person, or an automation you control, should send it. Check that each message names the room, the fault, the date it was first reported and the date you need it fixed. Vague chasers are easy to ignore.
Who should mark a repair as closed?
The person who did the work marks it fixed, and a second person confirms it before the request closes, usually housekeeping or the duty manager checking the room. For a room out of order, nobody puts it back on sale until that check is done, because a guest who finds the same fault becomes a complaint and a repeat request in the same week.
Further reads
- Hotel SOPs Written With AI: Check-In, Housekeeping, Night Audit — Write the inspection steps that catch faults before guests do.
- Hotel Guest Complaints Handling: Service Recovery Scripts With AI — What to say when a fault has already reached a guest.
- Hotel Guest Feedback: Turn Post-Stay Surveys Into Fixes With AI — Survey comments often point to the repeat faults in your log.
- AI Maintenance Costs: What You Pay After an Automation Goes Live — What running an AI routine costs after setup.
- What Software Do Hotels Use? PMS, Channel Managers and AI Add-Ons — How maintenance tools relate to your property system.
- AI Agents for Hotels: What They Can Do Today and Where They Fail — What a hotel AI agent can handle now, where it fails on bookings, guest messages and housekeeping, and the tests and hand-off rules to set first.
- Can AI Build a Gym Equipment Maintenance Checklist From Manuals? — How to turn gym equipment manuals into a checked maintenance checklist, a service calendar and a fault log with AI, and which jobs stay with qualified people.
- Smart Room Automation Through AI in Hotels: What a Small Property Can Add — What a small hotel can add to its rooms first: thermostat setback linked to the booking system, window and door sensors, and clear guest consent.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: MaintainX and UpKeep pricing pages and Flexkeeping's website (checked October 2026); vendor help pages on out-of-order and out-of-service room statuses. The test figures and revenue sums are illustrations.