How do you measure AI ROI?
By the InTheory team · Updated November 2026
You measure AI ROI by recording a baseline before any change, then comparing the same numbers after rollout against the full cost. Track time per task, turnaround, error or correction rate, and a revenue measure such as conversion rate. Include software, setup, training and review time in the cost, not just the subscription.
Build the baseline first
Most AI projects cannot show a return because nobody measured the starting point. Before changing a workflow, sample 10 to 20 recent examples. For proposals, record staff minutes per proposal, days from call to sent proposal, and how often numbers had to be corrected. For lead response, record median time to first reply and the share of enquiries that became booked appointments.
Count the whole cost and the whole gain
Costs include licenses, implementation fees, integration tools, staff training and the time people spend reviewing AI output. Gains include hours saved, fewer errors, faster cash collection and extra wins.
As an illustrative example, if an engagement costs $5,000 plus $500 in tools and produces five extra wins worth $4,000 each at a 40% margin, the contribution is $8,000, a net of $2,500. Break-even would be $13,750 in additional sales. This is arithmetic, not a forecast. InTheory compares results against the baseline between days 61 and 90.