How AI Triages Maintenance Requests Over the Phone
Triage is the function that makes AI genuinely useful in property management, not just as a call answering tool but as an operational system. Any voicemail can record that
a tenant called. Triage determines what the tenant called about, how serious it is, and what should happen next.
Without triage, every maintenance call is treated the same. A broken boiler in January and a scuffed skirting board both produce a message that says "tenant called about a
maintenance issue." The property manager or administrator then has to call back, re-collect the information, and make the urgency assessment themselves. Every call
becomes two interactions.
AI triage eliminates that redundancy. The AI assesses the urgency on the initial call, routes appropriately, and hands off a complete, categorized request. The property
manager receives an action-ready log, not a raw message.
The Three-Tier Triage Framework
Maintenance calls in property management fall into three categories that determine the response required.
Emergency. Issues that pose immediate risk to life, safety, or significant property damage. These cannot wait. They require immediate escalation to an on-call team
regardless of time.
Examples: fire or smoke, gas leak or smell, carbon monoxide detection, burst pipe causing flooding, total structural failure, security breach leaving the property
unsecured.
Urgent. Issues that significantly affect habitability but are not immediate safety risks. These require same-day or next-morning response, not next-available scheduling.
Examples: no heating during cold weather, no hot water, major plumbing fault affecting use of facilities, complete loss of electricity to a unit, broken locks on a primary
entrance.
Routine. Issues that are inconvenient but do not affect habitability or safety. These are logged and scheduled within the standard maintenance SLA, typically 24 to 72
hours for the first contact with the tenant.
Examples: broken appliances, dripping taps, damaged fixtures, minor damage to walls or floors, pest sightings without immediate health risk, issues with shared facilities.
Getting this categorization right on the initial call is the core of what AI triage does.
Step 1: Open Question
The AI's first substantive question after identifying the tenant and property is an open invitation to describe the issue. "Can you tell me what's happening?" or "What's
the problem you're calling about?"
The open question matters because tenants describe problems in their own language, not in maintenance categorization terminology. A tenant does not say "I have an urgent
plumbing fault." They say "there's water coming through the ceiling in my bathroom" or "the boiler is making a strange noise and there's no hot water."
The AI listens to this initial description and begins assessing urgency based on keywords and context. Certain words and phrases trigger immediate emergency assessment:
fire, flood, gas, smell of gas, smoke, carbon monoxide, no heat, ceiling collapse, broken front door.
Step 2: Clarifying Questions
For descriptions that are not immediately clear, the AI asks targeted clarifying questions to establish urgency and scope.
For heating and hot water issues:
"Is this affecting the whole property or just part of it?"
"When did this start?"
"Is there anyone vulnerable in the property, such as young children or elderly residents?"
For water-related issues:
"Is the water actively flowing or is it a drip?"
"Do you know where the source is?"
"Have you been able to turn off the water supply?"
For electrical issues:
"Is this a complete power failure or affecting specific circuits?"
"Have you checked the fuse board?"
"Are there any signs of burning or unusual smell?"
For security issues:
"Is the door currently unsecured?"
"Are you safe in the property right now?"
These questions are not generic. They are configured by the property manager based on the specific properties and typical fault types in their portfolio. An older building
with known issues around certain systems gets a question set tailored to those systems.
Step 3: Category Assignment
Based on the initial description and the responses to clarifying questions, the AI assigns a category. This assignment determines what happens next.
Emergency assignment triggers immediate escalation. The AI transfers the call to the on-call emergency number before completing any further intake. The full call
transcript is logged simultaneously.
Urgent assignment triggers a notification to the maintenance coordinator flagged as requiring same-day or next-morning response. The AI completes the full intake, provides
the tenant with a reference number and timeline, and closes the call.
Routine assignment triggers a standard maintenance request log. The AI provides the tenant with a reference number and an expected response timeline consistent with the
standard SLA for routine issues.
Step 4: Information Collection
For urgent and routine calls, after the category is assigned, the AI collects any remaining information needed for the maintenance request to be actionable.
The minimum required information for a complete maintenance log is:
The tenant's name and contact number, confirmed during identification. The property address and unit number. A description of the fault in the tenant's own words. The AI's
categorization and the basis for it. The time and date of the call. Any relevant context: how long the issue has been present, whether it is getting worse, whether access
has any constraints.
Some property management systems also benefit from additional information: whether the tenant is available for access and when, whether there are specific entry
instructions, or whether there is a preferred contractor the tenant has worked with before.
The AI collects this systematically. The result is a complete, structured log that the maintenance coordinator can act on without any follow-up call to the tenant.
Step 5: Tenant Confirmation
The AI closes the call by reading back the key details of the logged request and providing confirmation.
"I have logged a routine maintenance request for a leaking tap in the kitchen at your address. Your reference number is 3841. Our maintenance team will be in touch within
24 hours to arrange access. You will receive a confirmation message shortly."
This confirmation step is not bureaucratic. It serves two functions. First, it gives the tenant the information they need to hold the system accountable. If they have a
reference number and a promised timeline, they know what to expect and when to escalate if nothing happens. Second, it reduces repeat calls. A tenant who received clear
confirmation does not call back the next day to ask if anyone received their message.
How the Triage Log Is Used
The structured log produced by the AI triage process is the input to the maintenance workflow. The property manager or maintenance coordinator receives a clear list of
requests, already categorized, with all relevant information included.
Emergency requests triggered escalation during the call. By the time the coordinator reviews their morning log, the emergency has already been actioned.
Urgent requests are flagged visually and appear at the top of the queue. The coordinator sees immediately which requests need same-day response.
Routine requests are sorted chronologically and scheduled within the standard SLA. No call-back to re-collect information. No uncertainty about what the issue is or which
property it relates to.
Staffify AI produces this structured output for every maintenance call it handles at €0.22 per minute. For a portfolio of 100 units generating 100 maintenance calls per
month at an average of 4 minutes per call, the monthly triage cost is €88.
FAQ
How does AI triage maintenance calls without human involvement?
The AI uses keyword detection on the tenant's initial description combined with structured clarifying questions to categorize the issue as emergency, urgent, or routine.
Emergency issues trigger immediate escalation. Urgent and routine issues are fully logged during the call with all information the maintenance team needs to act.
What happens if the AI miscategorizes a maintenance call?
The AI is configured conservatively: when in doubt, it escalates rather than downgrades. A description that could be a gas issue is treated as an emergency until
clarifying questions confirm otherwise. Property managers review the category logic during setup and adjust thresholds to match their risk tolerance.
Can AI triage work for portfolios with many different property types?
Yes. The triage framework is configured per property or portfolio type. An older building with known boiler issues gets different clarifying questions than a modern
building with a different maintenance profile. The AI applies the right questions for the right property.
Does AI triage replace the need for a maintenance coordinator?
No. The AI handles the initial call intake and categorization. A maintenance coordinator still assigns contractors, manages access, and oversees resolution. The AI
eliminates the first-call redundancy and produces a complete, categorized log so the coordinator's work starts with action rather than information gathering.
What information does AI collect during a maintenance triage call?
The AI collects tenant identification, property details, a description of the fault, the urgency categorization and its basis, time of call, and relevant context such as
duration of the issue and access availability. The result is a complete structured log that is immediately actionable by the maintenance team.
See how Staffify handles your customer journey