The Challenge
The owner wasn’t short on demand, his ads were working and calls were coming in. The problem showed up when he actually pulled his call logs: 27 out of roughly 100 after-hours calls had gone completely unanswered. His techs are on roofs and under sinks all day. His office manager leaves at 5pm. After 6pm, every call either hit his personal cell, which he ignores during dinner, or fell straight to voicemail.
He had no idea how many of those missed calls were $400-$1,200 emergency jobs. What he did know was the psychology on the other end: a homeowner with a dead AC at 2am, or water pouring into a basement, isn’t waiting for a callback. They call the next name on Google, and a competitor books the job before he even wakes up.
He’d already priced a 24/7 human answering service and rejected it: $600-$1,500 a month for someone reading a script who can’t tell "my AC is making a weird noise" from a genuine emergency. Urgent calls would still land in a generic queue instead of getting flagged and routed immediately.
The Solution
We built a purpose-built emergency-triage voice agent, not a generic answering script:
- Semantic urgency classification - trained on trade-specific emergency signals ("flooding," "gas smell," "no heat below freezing") to detect real emergencies by meaning, not keywords.
- Structured warm handoffs - every transfer to a live tech includes a full context packet (name, issue, urgency tier, address, quoted symptoms) via SMS/CRM before the human ever picks up, so no caller repeats themselves.
- Full observability layer - transcript logging, confidence scoring, and real-time alerts the moment a call fails or hits low-confidence intent.
- Outcome-based reporting - built-in before/after analytics tracking calls answered, urgency classification accuracy, and jobs booked, so ROI is measured, not assumed.
The system was voice-tuned and stress-tested against real, messy caller behavior (accents, interruptions, background noise, angry customers) before it ever touched a live call.
The Result
Within one week, the after-hours gap that had been quietly costing the business $400-$1,200 per missed job was closed. Every after-hours call is now answered instantly, classified for urgency, and either resolved or hard-routed to a live tech with full context attached. The client confirmed satisfaction against every success criterion set at kickoff: voice quality, urgency detection, handoff quality, and cost predictability.
He no longer wonders how many emergency jobs went to a competitor last night. The agent already answered.