When everything is an emergency, nothing is
Walk into any property management office on a Monday morning and look at the maintenance queue. A burst pipe sits in the same undifferentiated list as a squeaky cabinet hinge. A resident with no heat in January waits behind someone who wants a lightbulb changed in a closet.
The problem isn't volume. It's that most maintenance intake treats every request as equal until a human reads it and decides otherwise. That human is busy, so the reading happens in batches, and the batch that included the gas smell got read forty minutes too late.
Triage is the discipline of sorting requests by severity the moment they arrive, before a coordinator ever touches them. Done well, it routes the real emergencies to a tech in minutes and lets the cosmetic stuff wait for the next scheduled window. Done poorly, or not at all, it turns your maintenance operation into a coin flip.
The four tiers that actually matter
Most property managers intuitively know there's a difference between "urgent" and "not urgent," but a binary split is too coarse. A four-tier model maps better to how repairs actually get scheduled and budgeted.
Tier 1: Life-safety emergencies
These threaten health, safety, or the structure itself. Gas leaks, electrical sparking or exposed wiring, flooding, sewage backups, no heat in freezing conditions, broken exterior locks on an occupied unit, fire or carbon monoxide alarms. These get dispatched immediately, day or night, no exceptions. The clock starts the second the request lands.
Tier 2: Habitability and urgent
Loss of an essential service that isn't immediately dangerous but makes the unit hard to live in. No hot water, a non-functioning refrigerator, a single inoperable toilet in a one-bathroom unit, AC failure during a heat advisory. These need same-day or next-morning attention, but they don't justify a 2 AM call-out.
Tier 3: Standard repairs
The bread and butter. A dripping faucet, a running toilet, a cabinet that won't close, a slow drain. These can be scheduled within a few business days and batched geographically to save trips.
Tier 4: Cosmetic and deferrable
Scuffed paint, a sticky window, a loose towel bar. Real requests that deserve a response, but they can wait for a planned visit or get bundled into turnover work.
The value of the tiers isn't the labels. It's that each tier carries a different response-time commitment and a different dispatch path, and the sorting happens automatically.
Why manual triage breaks down
The reason most operators don't triage well isn't ignorance. It's that the sorting work competes with everything else for human attention.
A resident describes their problem in their own words, often vaguely. "My kitchen is acting up." Someone has to read that, ask clarifying questions, decide what it means, and assign a priority. When the coordinator is on a call or out at a property, that decision waits. The request that needed a sixty-second judgment call instead aged in an inbox for two hours.
There's also the consistency problem. Two coordinators will tier the same request differently. One treats "AC isn't cooling well" as standard; the other, knowing a heat wave is coming, treats it as urgent. Neither is wrong, but the resident experience becomes a function of who happened to read the message.
And after hours, manual triage often doesn't happen at all. The request goes to voicemail or a generic form, and nobody assesses severity until the office opens.
What automated triage looks like
An AI intake agent reads every request the moment it arrives, on any channel, and applies your tiering rules consistently. When a resident texts "there's water coming through my ceiling," the system recognizes the flooding signal, classifies it as Tier 1, and routes it to the on-call dispatch path immediately, even at midnight.
When the description is ambiguous, the agent asks the same clarifying questions a good coordinator would. "Is the water actively spreading, or is it a stain that's been there a while?" The answer determines the tier. This back-and-forth happens in seconds, not after a callback the next morning.
The practical wins stack up:
- Consistent severity calls because the rules are applied the same way every time, regardless of who's working
- Round-the-clock coverage so a Tier 1 at 11 PM gets dispatched at 11 PM
- Cleaner queues because standard and cosmetic work is already separated from the urgent stuff before anyone opens the dashboard
- An audit trail showing when each request came in, how it was classified, and why, which matters when a habitability dispute lands on your desk
This is where Castellan's maintenance coordination fits. The AI agent handles first-contact triage across phone, SMS, and email, asks the diagnostic questions, assigns the tier, and only pulls in a human when a request genuinely needs judgment.
The cost math of getting triage wrong
Mis-tiering cuts both ways, and both directions cost money.
Over-escalate, and you pay emergency call-out rates for work that could have waited. A standard after-hours plumbing dispatch can run two to three times the daytime rate. If your team is sending Tier 3 drips out as emergencies because nobody sorted them, you're burning budget on premium labor for non-premium problems.
Under-escalate, and the cost is worse. A small active leak classified as "standard" and scheduled for Thursday becomes water damage, mold remediation, and a displaced resident by Wednesday. The Bureau of Labor Statistics and most insurance data point the same direction: water intrusion handled late is one of the most expensive categories of property loss there is.
Then there's the habitability liability. When a resident reports no heat and the request sits unsorted in a queue, you're not just risking a bad review. In many jurisdictions you're risking a code violation and a rent-withholding claim. Accurate, timely triage is part of your compliance posture, not just an operational nicety.
Building triage rules you can trust
Good automated triage starts with rules that reflect your portfolio, not a generic template.
- Define your tiers explicitly, with example phrasings residents actually use, not just clinical descriptions
- Set response-time commitments per tier and make them visible, so residents know what to expect and your team knows what it's accountable for
- Account for context like weather, unit type, and whether a failure leaves the resident with zero alternatives. A second bathroom out of service is standard; the only bathroom out of service is urgent
- Decide your escalation thresholds for when the AI should stop and hand off to a person, such as a request involving injury, a legal threat, or anything the system can't confidently classify
Then watch the edge cases. The first few weeks of any automated triage system will surface requests that don't fit cleanly. That's not a failure, it's tuning. Each ambiguous case you resolve becomes a sharper rule for the next one.
The payoff is a queue that sorts itself
The point of triage automation isn't to remove humans from maintenance. Your techs still fix things, and your coordinators still handle the calls that need a person. The point is to make sure the urgent work never waits behind the trivial work, and that the decision about which is which doesn't depend on who happened to be at their desk.
When triage runs automatically and consistently, emergencies get the speed they demand, routine work gets batched efficiently, and your team spends its energy on repairs instead of on sorting. That's the difference between a maintenance operation that reacts and one that runs.