Home Services & The Trades: what to automate, what to keep human.

Front-office AI in home services is a real revenue lever and the market knows it: Avoca raised over $125M at a $1B valuation in April 2026 to put AI on the phones for HVAC, plumbing, roofing, and electrical shops, and it is running for 800+ contractors. The demo earns the money. The AI answers every call, books the job, follows up on the estimate, and pushes it to dispatch with no CSR. The stakes are real too: as Avoca's own framing puts it, a restaurant that misses a call loses a $30 order; a home-services business that misses one can lose a $30,000 or $40,000 HVAC install.

Here is the part of the math the demo skips. Two different calls get bundled into one product. The call that reaches the AI and gets booked is the demo. The call that never reaches anything, the second line during a heat wave, the 9pm no-heat call, the one that rang four times while the AI was mid-conversation with someone else and rolled to a voicemail no one has owned for years, is the bigger leak by dollar value. Front-office AI sold without an owner for that second call gives you a busier dashboard and a quieter monthly leak.

So a serious owner's question is not 'can it book the call.' It is 'what happens to the calls it does not, and who sees them before the install walks down the street to the next shop.'

Front-office AI is real, and the lower-risk pieces should move first because the revenue is right there. Missed-call text-back so an unanswered call gets an immediate 'we saw you called, what is the address' instead of dying in voicemail. After-hours capture that takes the no-heat call at 11pm, qualifies the emergency, and routes it to whoever is actually on call. Estimate and review follow-up the AI drafts and a human sends. Appointment reminders that cut no-shows. None of these dispatch a crew or quietly close the loop on a job with no human ever seeing it, and each one converts revenue you were already losing.

These earn AI its place in booking and dispatch later, once a named owner on the dead-call list and a callback clock actually exist and the missed-call number is moving. Build the workflows where the gain is visible and the miss is cheap first, then earn the one where a miss is a five-figure job that left without a trace.

Trades-AI features look like one bundle: call answering, booking, estimate follow-up, dispatch, review requests, reminders. They do not fail the same way. A late review request costs a Google star. A clumsy reminder annoys a customer. A call routed slightly wrong gets fixed at the next dispatch huddle. A call that no human and no AI ever caught is a job that left and told you nothing, and that one has a five-figure ticket on it.

That asymmetry is why order matters. ServiceTitan's 2026 Residential State of the Trades Report (1,000 contractors, Thrive Analytics, April 2026) found 74% of contractors see AI as key to efficiency and 73% think starting early is an advantage, but only about a quarter are actually using it and more than half say they do not know where to start. The same report shows barely half of contractors respond to a lead within the first hour. The leak is already there, in the dark, before any AI is installed. The job of the AI is to put a number and an owner on it, not paper over it with a booked-call dashboard that looks busy.

Practically: before the AI books a job, a named person owns the dead-call list and a clock runs on it (every missed call gets a human callback inside a defined window, tracked, with a reason logged when it does not convert). AI handles volume. A person owns the ones that fell through, because an AI graded on calls answered will never surface the ones it did not.