Why this matters
Monday morning: ten maintenance requests from the weekend in a spreadsheet, in the tenants' own words. One is a gas smell, one is water through a ceiling, one is a dripping faucet. Reading them in the order they arrived means the gas smell waits behind the faucet.
ChatGPT sorts them in seconds, picks the trade, and drafts the tenant texts and vendor lists. It does not know your local rules, your vendors or what already happened over the weekend. You make the safety calls.
Practice
Do every step yourself in a practice copy of the real tool. It checks your work, and “Do it for me” finishes any step you get stuck on.
What you’ll build
You manage three buildings for Juniper Property Management (made up). You attach the weekend's ten requests and get a table: three Emergency, three Urgent, three Routine, one that needs a question first, each with a reason and a trade.
Then the next 15 minutes in order, a text for each Emergency and Urgent tenant that promises no time you have not given, and one dispatch list per trade with unit, phone and entry permission.
What you need
- Nothing to start the practiceThe lesson is a practice copy of ChatGPT. The tenants and buildings are made up.
- A ChatGPT account, to do it for realA plan that lets you attach a file. The answers shown came from a Plus account.
- The requests with their timesAn export from your request form or portal, including when each came in and the entry permission.
- Your emergency rulesWhat your company and local code treat as an emergency, such as no heat below a certain temperature.
Concepts
Sort, and say what is missing
The request asks for three levels, a reason in a few words and a trade, and "if a request doesn't have enough information to sort, say what to ask." The real run held the refrigerator request ("Something is wrong with the fridge", no phone) and listed the questions: is it cooling, any smell, smoke or water.
"Be careful with anything that could be a safety issue" got the gas request the right answer: leave, avoid switches and flames, call 911 or the gas utility from outside.
Read the times yourself
The gas smell came in Saturday at 11:42 PM. It is now Monday. ChatGPT never mentioned that. Your first call is to find out whether anyone already responded on Saturday night, not to send a text as if it just arrived.
A triage table sorts by what was reported. You know when, and what happened since.
Your rules, your call
No heat at 58 degrees with a four-month-old was ranked Urgent, not Emergency, with a note to escalate if it gets colder. Many managers, and many local housing codes, would call it an emergency now. Decide by your own rules.
The texts and dispatch lists are ready to send once the levels are yours. Keep its "Entry: Call first" lines: they are the tenant's instructions.
Do it for real
Now repeat it in your own account. Tick each step as you go.
Downloads
- maintenance-requests-monday.csvThe ten made-up requests used in the lesson.
Quick check
3 questions to check you’ve got it.
1. A tenant reports a gas smell. What is the first instruction?
- Send a plumber
- Leave the building and call 911 or the gas utility from outside
- Open a window and wait
Show the answer
Leave the building and call 911 or the gas utility from outside. A suspected gas leak is always leave first, then call from outside. Nothing else comes before it.
2. The gas report came in Saturday night and it is Monday. What do you check first?
- Whether anyone already responded over the weekend
- The tenant's lease
- The vendor's price
Show the answer
Whether anyone already responded over the weekend. The table shows what was reported, not what happened next. Find out before you act.
3. ChatGPT ranked no heat with a baby as Urgent. Your local code says it is an emergency. Which wins?
- ChatGPT's ranking
- Your local rules and your judgment
- Whichever is faster
Show the answer
Your local rules and your judgment. ChatGPT does not know your code or your building. The levels are yours to set.
Fillout: Build a Tenant Maintenance Request Form That Flags Emergencies
Triage is faster when requests arrive complete. Next, build the request form that asks for the details and flags emergencies on its own.
Start module 4 →