After-Hours Call Handling for Trades

The call that comes in after the office closes is the one that matters most

A furnace that dies at 4pm on a Friday in January does not wait until Monday morning. Neither does a burst pipe on a Saturday night or a commercial refrigeration failure that starts spoiling inventory the moment the compressor quits. For trades that carry genuine emergency work, the after-hours phone is not a nice-to-have. It is the difference between a customer who calls you back next season and one who found a 24-hour competitor because your line rang out.

Most small trades businesses solve this badly, in one of two ways: a voicemail box nobody checks until morning, or an owner or one technician personally fielding every call around the clock, including the ones that could safely wait until 9am. Neither is sustainable, and neither actually sorts urgent from routine, it just moves the burden around.

After-hours call handling is a system that answers every call outside your normal hours, decides whether it’s genuinely urgent against a list you defined, and routes accordingly: a real emergency reaches your on-call technician within minutes, and everything else lands as a clean, organized message queue waiting for your office when it opens.

The core design decision: a written emergency list, not a judgment call

The single most important part of this system isn’t the phone technology. It’s the list. Before any call routing gets built, you sit down and define, in writing, exactly what counts as urgent enough to wake someone up. For an HVAC contractor that might be: no heat with an outdoor temperature below a threshold, a gas smell, no power in freezing conditions, an active leak. For a plumber, it’s different: active flooding, no water at all, a sewage backup. The list is specific to your trade because “emergency” means something different to every one.

This matters because the alternative, an AI or a live agent making an ad hoc judgment call about whether a caller’s description sounds urgent, is exactly where these systems get things wrong. A defined trigger list removes the guesswork: certain words and situations always escalate, regardless of how calmly or anxiously the caller describes them, and everything else defaults to a next-business-day callback with the details captured cleanly.

How the escalation actually works

1. The call comes in after hours. The system answers immediately, no ring-out to voicemail.
2. It runs a short triage. A handful of questions matched against your defined emergency list, not an open-ended conversation.
3. If it matches an emergency trigger, it escalates immediately. Your on-call rotation gets called in order, with a defined wait time before it tries the next person, so a real emergency never just times out unanswered.
4. If it doesn’t match, it takes a detailed message. Name, number, address, and the issue in the caller’s own words, queued cleanly for your office at opening.
5. Every call is logged and recorded, both so you can audit how the system is performing and because a documented call trail matters if a dispute ever comes up about what a caller was told and when.

What this costs, across the three ways to build it

Live human answering services are the traditional option and remain a reasonable choice for many shops. Ruby Receptionist’s published tiers run $250/month for 50 minutes up to $1,725/month for 500 minutes; Smith.ai runs $300/month for 30 calls up to $2,100/month for 300 calls, with per-call overage between $8.50 and $11.50 (both checked 2026-08-28). You’re paying for a live person handling the call and the judgment calls that come with it, at a scale of cost that climbs quickly with volume.

Packaged AI answering products cost less. aiphonereceptionist.ca prices around $125/month including 50 minutes, then $0.25/minute after that (checked 2026-08-28), running a fixed script template rather than a rule set custom-built for your trade’s emergency definitions.

A custom-built system, wired to your specific escalation list and on-call rotation, runs on raw per-minute voice-platform costs rather than a subscription: Retell AI’s published components put a typical call in the range of $0.07 to $0.31 per minute depending on the language model and voice chosen (checked 2026-08-28). There’s no monthly minimum, but there is real cost in building and maintaining the escalation logic, and that cost doesn’t disappear just because it isn’t a recurring line item.

Where a human backstop is not optional

However this gets built, one design rule doesn’t change: anything ambiguous should escalate to a human, not get resolved by a script’s best guess. A false alarm costs an on-call technician a disrupted night. A missed real emergency, dismissed because a caller described a gas smell calmly instead of frantically, is a genuinely dangerous failure mode, not an inconvenience. Bias the trigger list toward over-escalation, review it regularly against what actually came in, and never treat “the system will figure it out” as an acceptable answer for anything safety-related.

That conservative bias is also why this is worth building as its own narrow system rather than folding it silently into a general AI receptionist: after-hours emergency triage deserves its own explicit rules, tested against your trade’s actual emergencies, not a general-purpose script doing its best.