An AI answering service for property management picks up every resident call, sorts real emergencies from routine requests using rules you set, pages your on-call person for the emergencies and turns everything else into a complete ticket. The AI should run the sorting and the paperwork. A person should still pick up when a pipe bursts.
That last part isn't me being cautious. Residents say it too. When HappyCo surveyed about 300 North American renters in 2025, 82% said they want a human first for an urgent maintenance problem, while 53% were comfortable using AI for routine requests (Multifamily Executive). So the job is to set the AI up to take the routine load and get urgent calls to a person faster than your voicemail does today.
Quick answer
- What the AI does well: answers every call, asks the right questions, sorts emergency vs. urgent vs. routine, pages on-call, writes the ticket with unit, problem, photos and access notes.
- What stays human: handling the emergency itself, lease decisions, fees and charge-backs, anything the agent can't classify.
- Where to start: after-hours calls for one portfolio, with an emergency list your maintenance lead signs off on.
- The setting that matters most: the escalation ladder. Who gets paged, how long before the next person, and what happens if nobody answers.
Why do after-hours calls cause so much trouble?
Because most of them aren't emergencies, and the few that are look the same in a voicemail inbox.
A typical night: a dripping kitchen tap, a noisy neighbour, someone locked out of the gym, a parking question, and one call at 2 a.m. about water coming through a ceiling. If every call goes to one on-call phone, your maintenance lead gets woken up five times. If every call goes to voicemail, the ceiling waits until 8 a.m.
Speed matters even for the routine ones. Buildium's 2026 Renters Report (1,500 US renters) found 84% expect a reply within 24 hours, and renters happy with upkeep were 57% more likely to renew (Buildium). A ticket logged at night with the right details is how you hit that without making staff answer at night.
What should count as an emergency?
Write this down before you buy or build anything. The AI follows the list. It shouldn't invent one.
Here's a starting table. Change it for your buildings, your climate and your local rules.
| Tier | Examples | What the agent does | Who acts and when |
|---|---|---|---|
| Emergency | Active water leak or flooding, no heat in cold weather, sewer backup, gas smell, fire or CO alarm going off, total power loss in a unit, broken exterior door or window, lockout with a child or medication inside | Gas, fire, CO and medical: tells the caller to call 911 first. Then pages on-call straight away, stays on the line to collect unit, details and access | On-call person, now |
| Urgent | No hot water, fridge not working, only toilet clogged, one breaker tripped, AC out in extreme heat | Logs a priority ticket, texts on-call a summary, tells the resident when to expect contact | First thing next morning, or on-call's call |
| Routine | Dripping tap, light out, appliance noise, pest sighting, cabinet door, parking and amenity questions | Logs a ticket with photos and access permission, answers FAQs from your building info | Normal work order queue |
| Not maintenance | Rent questions, lease renewals, noise complaints, package issues | Takes a message for the office, answers what your FAQ allows | Office, next business day |
Two rules I'd add to every list. First, if the caller sounds panicked or says "emergency" and the agent isn't sure, it escalates. A false alarm costs one wake-up. A missed flood costs a ceiling. Second, seasons change the tiers. No heat in July is urgent at most. No heat in January is an emergency.
How should the escalation ladder work?
This is the part generic answering setups skip, and it's the part your on-call person cares about most.
- Page the primary on-call by call and text, with a one-line summary: unit, problem, how long it's been happening.
- Wait a set time (we usually suggest 5 to 10 minutes) for them to acknowledge.
- Page the backup if there's no answer. Then the property manager. Then the owner of the company, if that's your rule.
- Call the resident back with what's happening: "Our on-call technician has your request and will call you within 20 minutes." Only say a time your team actually commits to.
- Log every step with timestamps: when the call came in, who was paged, who answered, when it was resolved.
That log is worth more than it looks. When an owner asks why a unit flooded, or a resident disputes a charge, you have the timeline. In places with legal repair deadlines, like Awaab's Law for social landlords in England, the timestamps are your evidence.
What does a good call sound like?
Here's a script you can adapt. It's short on purpose. The agent's job is to collect the right details fast, not to chat.
Agent: Thanks for calling Maple Street Properties. I'm the after-hours assistant, an AI. If this is a fire, gas smell or medical emergency, please hang up and call 911. Otherwise, what's going on?
Resident: There's water dripping from my bathroom ceiling.
Agent: Okay. Is it dripping, or is water running steadily?
Resident: It's getting worse.
Agent: Thanks. I'm treating this as an emergency and paging our on-call technician now. What's your unit number?
Resident: 304.
Agent: Unit 304. If you can, put a bucket under it and move anything electrical away from the water. Is it okay for the technician to enter if you're not home?
Resident: Yes.
Agent: Got it. The technician will call you at this number. If you don't hear back in 20 minutes, call this line again and I'll page the backup.
Notice what it does: says it's an AI up front, gives the 911 line first, asks one question at a time, confirms the unit number back, collects entry permission, and gives a next step with a time. Everything said goes into the ticket.
Example calculation: what triage saves your on-call person
These are example inputs, not benchmarks. Use your own call log.
| Input | Example |
|---|---|
| After-hours calls a month | 120 |
| Share that are true emergencies | 10% (12 calls) |
| Wake-ups today (every call goes to on-call) | 120 |
| Wake-ups with triage (emergencies only) | 12 |
| Routine calls turned into tickets overnight | 108 |
The point isn't the exact number. It's that your on-call person stops answering for light bulbs, so they're awake and quick when the ceiling call comes in. People quit on-call rotations over the light-bulb calls, not the floods.
Should you buy a tool or have one built?
It depends on how specific your rules are.
| Your situation | Good fit |
|---|---|
| Under about 100 units, one software system, simple emergency list | Your property software's add-on or a self-serve AI receptionist |
| Live answering already in place and you just need overflow | Keep the live service; add AI for overnight intake only |
| Several properties, your own forms, a co-op board, an on-call rotation with backups | A custom build around your rules and systems |
| Several systems that don't talk (phones, work orders, accounting, email) | Custom, because the value is in connecting them |
Whatever you pick, ask the vendor to show you three things in a demo: your emergency list driving the decisions, the escalation ladder with a no-answer fallback, and a ticket landing in your work order system with all the fields filled.
What should you ask a vendor before signing?
- Can we set our own emergency list, and change it by season?
- What happens if on-call doesn't answer? Show me the fallback.
- Does the agent tell callers it's an AI, and can they ask for a person?
- Does it write tickets straight into our system, or just send emails?
- Can we hear every call and read every transcript?
- Does it handle Spanish or French if our residents need it?
- Who owns the call recordings and data, and can we export them?
If a vendor can't answer the second question clearly, don't put them on your emergency line.
Where Trilix fits
We build phone agents and maintenance intake around a property manager's own rules, and we've built digital inspections for an Ontario housing co-op (case study). If you want the full picture of what we do for property teams, see AI for property management. For how AI receptionists compare across industries and price points, see our AI receptionist guide.
Want to see how triage would work on your call log?
Send us a week of after-hours calls and we'll show how they'd sort into emergency, urgent and routine.