Measure the baseline before any engine goes in

McKinsey's late-2025 survey found only 7% of organisations had scaled AI into results across the business. Ask the other 93% what their numbers were before the pilot and most cannot say. Without a baseline there is no result, only an impression, and impressions do not survive the first invoice. The baseline is the cheapest part of the whole project and the part most often skipped.

The phone

Total inbound calls, calls answered by a person, time to answer, calls after hours, calls that became bookings. The 411 Locals study of small businesses found 62% of calls not answered by a person; your phone system knows your own figure and has known it for years.

Leads

Time to first response, contact rate, booked meetings or consultations per inquiry, by source. Harvard Business Review's audit of 2,241 companies found an average response of 42 hours; measure yours before anyone promises to beat it.

The back office

Hours per task for the repetitive work: intake paperwork, document chasing, scheduling, reporting, data moved between systems. Turnaround time for each. Error and rework rates where you have them.

Revenue

Pipeline created, win rate, cycle length, average ticket, no-show or cancellation rate. Where the number comes from and who owns it.

How to get it in a week

Phone logs from your provider, a CRM or spreadsheet export, the calendar, the practice or job management system, and one week of a staff member noting hours on the repetitive tasks. Put it on one page with the date. That page is the contract you hold every vendor to, including us, and the only evidence that a result was a result.

The trap

Never accept a vendor-supplied baseline. If the same party sets the starting line and reports the finish, the race is theatre. Measure it yourself, or have the vendor measure it from your systems with you watching, before anything goes live.