Answer · Home services
How should a home services business use AI on the phone?
To catch the calls nobody could answer and to capture the job details, not to hold the conversation the customer wanted.
To answer the calls that would otherwise ring out, take the details, and call back or text within minutes. The gain is against a missed call rather than against a person answering, which is the comparison that makes it worthwhile.
The comparison determines the answer. Against a competent person answering, an automated system is worse on every axis that matters: it cannot judge urgency the way someone who knows the trade can, it cannot reassure, and it cannot make the small commercial decisions that turn a call into a job. Against a call that rings out while everyone is on a roof, it is enormously better, because the alternative outcome is the customer calling the next company.
So the design should target the second situation specifically. Out of hours, during jobs, when the office line is engaged, when everyone is on site. In those windows the system answers, establishes what is needed and where, records it, and either books something or promises a call back within a stated time. Nothing about that requires it to handle a complex conversation, and attempting one is where these systems produce the experiences people complain about.
Urgency routing is the part that has to work and the part that carries the risk. A water leak, no heating in winter, an electrical smell, a lockout: these need a person now, and a system that treats them as ordinary enquiries has done real damage. The reliable construction is a broad rule — anything mentioning water, gas, smoke, electricity, heat or a safety word escalates immediately — rather than a nuanced judgement about severity, because the cost of a missed escalation is concentrated in exactly these calls.
Capture is the second job and it is where the routine value sits. Name, number, address, what the problem is, how long it has been happening, whether it is a repeat, whether they are an existing customer. These are the fields whoever handles the job needs, they are the ones that otherwise get scribbled and lost, and having them in a record means the callback starts from information rather than from asking again.
The callback promise is the whole commercial mechanism and it has to be kept. A system that says someone will call within the hour and nobody does has produced a worse outcome than the missed call, because it added a broken commitment to it. Whatever window is stated must be one the business actually meets on its worst day, which is usually longer than anyone wants to say and is still much better than silence.
Recording and consent are worth settling before anything is deployed. Rules on recording calls vary by state and some require all parties to consent, so the announcement and its wording are a legal question rather than a courtesy. This applies to transcription as much as to recording, and it is cheaper to resolve at setup than after a complaint.
Compare it against the call nobody picked up, not against your best person on a quiet morning.
Siddharth Sharma, Context Theory
Related questions
Will customers hang up on an automated answer?
Some will, and the relevant comparison is still what happens now. A caller who hangs up on a system was a caller who would have reached a voicemail or a ringing tone, and both of those also end the call. What reduces hang-ups is being brief, being clear that the details are being taken, and stating when a person will call.
Should it try to book the job?
For standard work against real availability, yes, and it is a genuine improvement over a callback. For anything where the duration or the parts are uncertain, a booking made without that knowledge produces a slot that cannot be worked, which is worse than a call back with an accurate time. Book the predictable, promise the rest.
METHOD
Every figure below carries its source and the date it was verified. Nothing on this page is asserted.
The numbers on this page.
| What | Value | Specific to |
|---|---|---|
| Firms that never responded to a web enquiry at all | 23% | Category-wide |
| Leads cold past 5 minutes | 93% | Category-wide |
Oldroyd, McElheran & Elkington, "The Short Life of Online Sales Leads", Harvard Business Review (March 2011) · 1.25M inbound leads across 2,241 US firms · verified
2026 speed-to-lead benchmark · derived: 100% − 7% responding within five minutes · verified
What is specific to this page.
| Kind | Claim | Check it against |
|---|---|---|
| Response | The relevant comparison for an automated answer is a call that would otherwise ring out rather than a person answering, because the alternative outcome in the targeted windows is the caller contacting the next company. | Counting calls that currently go unanswered during working hours, out of hours and while lines are engaged. |
| Workflow | Urgency escalation should use a broad keyword rule covering water, gas, smoke, electricity, heat and safety terms rather than a judgement about severity, because the cost of a missed escalation is concentrated in precisely those calls. | Testing the configured system with an emergency description and observing whether it escalates or takes details. |
| Constraint | A stated callback window that the business does not meet produces a worse outcome than the missed call, because it adds a broken commitment, so the window must be one achievable on the worst day rather than the typical one. | Comparing promised callback times against actual callback times over a busy period. |
| Regulation | Call recording rules vary by state and some require the consent of all parties, so the announcement and its wording are a legal requirement rather than a courtesy, and the same applies to transcription. | The recording consent statute of the state or states in which the business operates. |
Each row would be wrong on another industry's page. Where a sourced figure exists it is in the table above instead; these are the constraints that shape the work and do not happen to be numbers.
Start with the measurement.
Reading about a benchmark is not the same as knowing your own number. The audit produces yours, measured rather than estimated.
$497 · delivered in 5 business days · credited against month one