Before you buy an AI receptionist, audit 7 days of missed calls

An AI receptionist should not be priced before somebody checks which calls the business is actually losing.

It is easy to build a convincing estimate with an assumed missed-call rate, an average job value, and a generous conversion rate. The result may look impressive, but it says very little about one specific HVAC or plumbing company.

A better starting point is seven days of call history. One week will not settle the business case, especially outside the busy season, but it can reveal when calls are being missed, what those callers wanted, and whether a focused pilot is worth testing.

Download the seven-day missed-call audit worksheet. Open it in Excel, Google Sheets, or Numbers and replace the example row with your own calls.

Record the call before estimating its value

For each inbound call, capture only the facts you can verify:

• date and time
• whether somebody answered
• whether the call came during business hours
• new enquiry, existing customer, supplier, recruitment, or spam
• what the caller needed
• when the team called back
• whether contact was restored
• whether the enquiry produced a booking
• the final job outcome, if known

Do not assign revenue to every missed call. An existing customer checking an arrival time, a supplier, a robocall, and a new emergency request have different value.

The useful number is relevant missed enquiries: calls from potential customers that the business wanted to answer but did not.

Separate after-hours calls from overflow

Two companies can miss the same number of calls and need different solutions.

If most opportunities arrive after closing time, start with after-hours coverage. If calls are missed while dispatch is already handling another customer, test overflow. If the calls require estimates, diagnosis, negotiation, or sensitive judgment, a human callback may remain the right outcome.

This distinction keeps the first pilot small. The agent receives a defined job instead of being asked to replace the entire front desk on day one.

Use contribution profit, not the invoice total

Suppose the audit finds eight relevant missed enquiries. That does not mean eight lost jobs. Some callers would not qualify, some would not answer a callback, and some would choose another company for reasons unrelated to response time.

Use conservative numbers from the business:

recoverable opportunities
= relevant missed enquiries × realistic contact-recovery rate
 
expected completed jobs
= recoverable opportunities × normal close rate
 
estimated contribution profit
= expected completed jobs × profit per completed job

Profit per completed job is more useful than revenue per job because labour, equipment, parts, travel, and other delivery costs still exist.

If those numbers are unavailable, keep the result as a range. A range built from known uncertainty is more useful than a precise number built from guesses.

Include the cost of follow-up and cleanup

Software cost is only part of the operating cost. A pilot may also require:

• phone and usage charges
• setup and integration work
• staff time reviewing uncertain calls
• corrections to customer or dispatch records
• human coverage for escalations
• ongoing testing when business rules change

The comparison becomes:

estimated monthly value
= contribution profit from recovered jobs
− system cost
− staff follow-up and cleanup cost

This prevents a common mistake: counting every answered call as a win while ignoring the work created by incorrect details, duplicate bookings, or poor handoffs.

Decide what the first pilot is allowed to do

The audit should lead to one of three decisions.

Keep the current process

If valuable calls are rarely missed, or the team already returns them quickly and consistently, another answering system may add complexity without solving a meaningful problem.

Test after-hours coverage

Use this when a clear group of valuable calls arrives outside working hours. The first version might answer common questions, collect a structured request, apply approved urgency rules, and create a callback task.

Test overflow coverage

Use this when the front desk or dispatcher loses calls during predictable rushes. The agent can capture the caller and reason for calling while the team remains responsible for complex decisions.

Direct booking should be included only when the scheduling rules and integration can support it safely.

Measure the record produced by the call

A pleasant conversation is not enough. During the pilot, inspect what reached the calendar, CRM, inbox, or dispatch queue.

Track:

• relevant calls answered
• qualified requests captured
• confirmed bookings
• successful transfers
• callback tasks created
• incorrect customer details
• duplicate or uncertain bookings
• callers who abandoned the conversation
• staff time spent fixing records

Compare those outcomes with the seven-day baseline. The useful question is not whether the voice sounded human. It is whether the business recovered worthwhile work without creating a larger operational problem.

Start with the numbers you already have

Seven days is a diagnostic, not a forecast for the whole year. Seasonal businesses should repeat the audit during a busier period before making a larger commitment.

Still, one careful week is enough to replace a vague sales claim with a specific operating question: which missed calls are worth recovering, and what is the smallest reliable workflow that could recover them?

Download the worksheet, fill it with fictional or redacted caller information, and calculate a conservative range. If you want a second pair of eyes on the result, bring the numbers to a short missed-call audit.