How do I measure ROI on an AI receptionist?

The circulating figures in this category are worth ignoring. Savings claims of tens of thousands a year trace back to vendor blog posts with no source behind them.

Your own numbers are better and take about twenty minutes to gather.

The calculation is simple enough to do on paper, which is a point in its favor.

First, missed calls a month. Your carrier's log has this, and counting one week by hand works if it does not.

Second, your conversion rate on inquiries. Most owners know this roughly, and roughly is good enough here.

Third, what a customer is worth over the relationship rather than on the first job. A repeat customer is worth several times the first invoice.

Multiply the three and compare against the plan price. Twenty missed calls, one in five converting, $400 a customer works out at $1,600 a month against $49.

Then keep measuring after you switch it on, because the estimate is a reason to try and the transcripts are the evidence.

An estimate tells you whether to try. Measuring afterward tells you whether to stay. Say you run a three-van electrical firm, and in month one the assistant answered 140 calls on no-answer forwarding. You spend an hour tagging the transcripts.

New-customer inquiries come to 38. Existing customers account for 61, vendors and junk for 22, and 19 are miscellaneous. Of the 38 new inquiries, 17 booked an estimate during the call and 9 turned into jobs averaging $620, which is $5,580 of work.

It's tempting to set $5,580 against a $49 subscription and stop. The fair comparison is against what you'd have won anyway. Before, those 38 callers hit voicemail, maybe half left a message, and you won about a third of the ones you called back. That's roughly 6 jobs compared with 9. The assistant's real contribution is about 3 jobs, or $1,860, which is still a strong month.

An honest sheet has a cost side with more than the subscription on it. Add any overage. Include the automation tool's monthly fee if you're paying for one only because of this. Your own reading time counts too: twenty minutes a day in the first fortnight and perhaps an hour a week after that, priced at whatever your time is worth.

Give bad calls their own column. Read the transcripts where a new inquiry didn't book and ask whether a person would have saved it. If three callers a month hung up at the greeting, that's a cost, and it belongs on the sheet beside the wins.

Be careful with staff time saved. It only counts if it turned into something, such as fewer overtime hours, invoices sent sooner or a front desk that stopped falling behind.

Judge a full month at minimum, and compare against the same month last year, since a busy season flatters any new tool that happens to start in it. Keep the sheet going for a quarter.

Pricing is published so the cost side of that sum is knowable before you talk to anyone: $49 for 200 calls, $149 for 750, $349 for 2,000.

Overage rates are printed too, at $0.45, $0.35 and $0.25 a call, which is what most vendors in this category leave off the page.

Transcripts let you check the benefit side honestly, including how many of the recovered calls were worth having.