Anonymised, illustrative composite. A six-building Ontario portfolio was paying premium after-hours callout rates for problems that were never actually emergencies — until it wrote down what counts as one.
At a glance
A single on-call coordinator answered every after-hours call across six buildings, and the standing instruction was simple: if a resident calls after hours, dispatch a vendor. That rule protected against the worst case — a missed genuine emergency — by treating every call as one, which is expensive in a completely different way.
Premium after-hours callout rates from vendors ran two to three times the daytime rate across the portfolio’s trades. Nobody had actually measured what fraction of those calls needed that premium response.
A four-week log of every after-hours call, tagged by what the resident actually reported, put a number on it: 62 calls, of which 41 — about two in three — were not emergencies by any reasonable reading. A flickering hallway light and a lukewarm-but-not-cold thermostat complaint do not carry the same urgency as a suite fire alarm or a burst supply line, but the dispatch process was not distinguishing between them.
The gap was not that the portfolio lacked a plan. CAO’s own description of the role confirms that responding to emergencies is a defined part of the job — the gap was that nothing in the plan actually said what an emergency was, so “respond to emergencies” had drifted into “respond to every call.”
Sorting the 41 non-emergency calls by category showed the same handful of issues recurring: 14 were minor plumbing (slow drains, running toilets), 11 were lighting or minor electrical, 9 were HVAC comfort complaints with no equipment failure, and 7 were miscellaneous items a resident could reasonably expect to wait until the next business day.
None of those 41 calls involved life safety, active water intrusion, no heat in cold weather, or loss of a vital building system — the kind of thing a genuine emergency plan exists to catch. Dispatching a premium-rate vendor to all of them anyway was not caution; it was the absence of a rule.
The 21 calls that did meet the emergency bar split roughly evenly across the six buildings, which mattered for the fix: this was not one problem property skewing the average, it was a portfolio-wide gap in how every call was being triaged at intake, regardless of which building it came from.
A documented emergency management plan is the thing CAO names as the manager’s actual duty — and a plan, by definition, has to define its own scope. The firm wrote a one-page triage list: named categories that always dispatch immediately (fire/life safety, active flooding, no heat below a stated outdoor temperature, loss of building power or elevators trapping someone), and everything else routed to a next-business-day queue with a same-day callback. Deferring a non-emergency under a documented plan that says so is not a lapse in the duty to respond to emergencies — it is what having a plan is for.
Measured over the following two months against the same four-week baseline, after-hours callouts fell from an average of 62 a month to about 41 — a reduction of roughly a third — without a single genuine emergency going unanswered.
The premium after-hours callout rate itself never changed, and the vendors on that rate were never asked to discount it. What changed was how often the portfolio actually paid it — from roughly 62 premium-rate dispatches a month down to about 41, with the other 21 or so resolved the next business day at ordinary daytime rates instead.
The triage list itself became the artifact residents were told about at move-in: which categories page the on-call coordinator immediately, and which get a callback the next business day. For the maintenance-request system this triage logic plugs into, see building a maintenance request system that triages itself, and for how the same portfolio schedules the preventive work that keeps emergency calls from happening in the first place, see AI-assisted preventive maintenance scheduling.
A 30-minute call is enough to tell you whether AI pays for itself here.
