How to triage emergency calls without a night dispatcher
How to configure an AI phone system to recognize true emergencies, route them to your on-call team, and avoid waking techs for non-urgent calls — across HVAC, plumbing, electrical, dental, locksmith, and other emergency-driven trades.
For trades and practices where calls happen 24/7 — HVAC, plumbing, electrical, locksmith, dental, veterinary, septic, garage door — emergency triage is the highest-leverage capability of any phone system. Done well, it captures the high-value emergency calls that competitors miss while protecting your team from getting woken up for non-urgent issues. Here is how to configure it.
The triage problem
Emergency-driven trades face two opposite failure modes. Mode 1: you have no after-hours coverage, so genuine emergencies (burst pipes, no heat in winter, lockouts at 2am) hit voicemail and book with whoever picks up. Mode 2: you have on-call rotation but no triage, so a tech gets paged at midnight for a "I have a question about my bill" call. Both fail; the first costs you revenue, the second burns out your team.
AI triage solves this by recognizing what is genuinely an emergency from the caller's opening words and routing accordingly. Real emergencies dispatch immediately; routine calls schedule into the next-available business-hours slot.
How to configure emergency rules per trade
During onboarding, you define what counts as an emergency for your specific business. The AI uses these rules to triage every inbound call in real time. Examples by trade:
- HVAC: "no cool" with house temperature over 80°F + occupants present, "no heat" with house temperature under 50°F or vulnerable occupants (elderly, infants), CO alarm sounding, gas smell, electrical burning smell from a unit.
- Plumbing: active flooding or sewage backup, no water (whole-house or critical fixture), gas leak suspected, broken sewer line, water heater leak with no shutoff.
- Electrical: burning smell from outlet or panel, sparking, smoke from electrical, tripped breaker that will not reset, partial loss of power affecting critical systems (medical equipment, refrigeration, well pump).
- Locksmith: stranded with kids or pets inside, locked out in extreme weather, after-hours business lock change required for security (employee fired today, etc.), keys lost late at night with safety concern.
- Dental: severe tooth pain with swelling, knocked-out tooth (90-minute reimplantation window), broken tooth with sharp edge causing oral injury, post-op bleeding that will not stop.
- Veterinary: pet not breathing or unconscious, suspected toxin ingestion, severe vomiting / lethargy with other symptoms, hit by car, post-surgical complications, severe difficulty breathing.
- Septic & sewer: active sewage backup in living spaces, drain field overflow, sewage in yard, multiple drains backing up at once.
- Garage door: stuck door with car inside (customer needs to leave for work), door stuck open at end of day (security concern), broken spring with car blocked.
The triage decision flow
Every inbound call goes through this flow on platforms with proper emergency triage:
- 1Caller speaks. AI listens to the opening message and qualifying answers.
- 2AI asks clarifying questions if needed. For ambiguous calls ("there's water on the floor"), the AI asks the right qualifying question ("is the water still actively flowing?") to determine emergency status.
- 3Triage decision. If the call matches an emergency rule for your business, route to emergency dispatch. If routine, route to normal scheduling. If genuinely ambiguous, route per your fallback rule (default: treat as emergency to be safe).
- 4Emergency calls are dispatched. Configurable: text/page your on-call tech with full caller context (address, issue, urgency cues), book a same-day emergency appointment, or transfer to a designated on-call number.
- 5Routine calls are scheduled. Booked into the next-available business-hours slot using your normal scheduling rules. Customer hangs up with a confirmation text.
False-positive vs false-negative tradeoff
Triage configuration is a tradeoff. Tighter rules (fewer calls flagged as emergencies) protect your team but risk missing the occasional real emergency that did not fit the pattern. Looser rules catch more emergencies but wake your team for false alarms. The honest recommendation: start slightly looser, log every emergency-flagged call for the first 2-4 weeks, then tighten based on what you actually saw.
What to send your on-call tech
When the AI dispatches an emergency, the alert your tech receives should include enough context that they can roll without a follow-up call. Configure your alert template to include:
- Caller name and callback number.
- Address (street + city; flag if outside service area).
- Issue description in the caller's own words.
- Severity / qualifying detail captured during the call (system age, symptoms, duration of issue).
- Whether the AI gave the caller any pre-arrival guidance (e.g. "walk them through finding the main shutoff" for plumbing).
- ETA window the AI told the caller — your tech needs to honor this or call back if they cannot.
After-hours rules vs business-hours rules
Triage rules can be different at night vs during the day. Some businesses configure them this way:
- Business hours: most calls handled normally; only true emergencies (active flooding, fire risk, life safety) trigger emergency dispatch.
- After hours: tighter rules — only "we cannot wait until morning" emergencies (burst pipe, no heat in winter, locked out with kids inside) get on-call paging. Other calls schedule for first thing in the morning.
This protects your team from after-hours pages for issues that genuinely can wait until 7am while still capturing the true emergencies.
Common questions
What if the AI mis-classifies a call?
Configurable backstops: you can require a confirmation step before emergency dispatch ("the AI thinks this is an emergency — call your on-call tech?"), have the AI default to "treat as emergency" for ambiguous calls, or set a maximum number of overnight pages per shift before falling back to next-morning scheduling. Most shops use a "default to emergency, log for review" pattern early on, then tighten.
Can the AI walk a caller through stopping the immediate damage (shutoff valve, etc.)?
Yes — and this is high-leverage. For active plumbing leaks, the AI can be configured to walk the caller through finding their main shutoff while the technician is dispatched. For electrical fire risks, walk through turning off the breaker. For pet emergencies, give pre-arrival instructions. These small interventions often save thousands in additional damage and build enormous customer trust during the worst moment of their week.
How do I handle calls where the customer is panicking?
AI receptionists configured for emergency-heavy trades use calmer pacing, shorter sentences, and reassurance phrases ("I understand this is stressful — let me get someone to you"). The conversation feels different from a routine call. If your platform allows tone adjustment, set the emergency-call tone to "calm and reassuring."
What about calls outside my service area?
Configure the AI with your service area boundaries (zip codes, cities, or radius from a base address). Out-of-area emergency callers should be politely told they are outside coverage and offered a referral to a partner shop in their area, not dispatched. This prevents accidental cross-state emergency calls and respects your tech's drive time.
Can I have different rules for different days or seasons?
Some platforms support time/season-based triage (e.g. "no heat" is an emergency in December but not in July). If yours does not, the workaround is to manually adjust rules at the seasonal change. Most owners do this once or twice a year and find it acceptable.
How to set up an AI phone answering system
Emergency triage is one of the highest-leverage configuration steps in the standard setup.
How to migrate from voicemail to AI answering
For shops currently using voicemail-with-callback, this is the upgrade path.
Related articles
How to set up an AI phone answering system in 30 minutes
A practical step-by-step guide for getting an AI receptionist live on your business phone in 30 minutes — from sign-up through the first answered call.
How to migrate from voicemail to AI answering
Why voicemail is functionally equivalent to no coverage, how to switch cleanly to an AI receptionist, and the parallel-running approach that lets you validate before committing.
Still stuck? Email support — we usually respond within a business day.