What does Inside Dr. Jonah Tebaa's Method for Measuring AI ROI That Isn't Just Hours Saved mean in practice?
Dr. Jonah Tebaa measures AI ROI through a capacity attribution ledger rather than hours saved. Instead of calculating theoretical wage savings, this framework requires leaders to decide in writing before rollout whether freed capacity targets revenue generation or cost avoidance. Value is only counted once it lands on financial statements as booked revenue from newly assigned tasks or as formally removed budget lines from avoided hiring.
Dr. Jonah Tebaa keeps encountering the same slide in AI project reviews: a big, satisfying hours-saved number, presented as if it were the return on investment. In his work advising teams on measuring AI value, he has stopped accepting that number at face value, and he argues most organizations should too.
The problem, as he frames it, is not that the hours are fake. The problem is what happens - or does not happen - after the hours are freed.
The Calculation That Looks Like ROI
Tebaa illustrates the issue with a generic scenario: a mid-size industrial parts distributor with eight sales reps, each drafting four customer quotes a day. Before an AI drafting tool, each quote took 25 minutes; after, with a rep still reviewing and approving every quote, it took 10 minutes. Fifteen minutes saved per quote, four quotes a day, works out to one hour saved per rep per day. Across eight reps and 22 working days a month, that is 176 hours.
Multiplied by a blended loaded wage of $14 an hour, the total comes to $2,464 a month - the number that shows up in the celebratory slide. Tebaa's point is that this figure never touches a real financial statement. No payroll shrank. No expense line changed. It measures time freed, not value created, and in his view, treating the two as interchangeable is where most AI ROI reporting goes wrong.
What Turns Freed Time Into Real Money
In Tebaa's account of the same scenario, the distributor's leadership did not wait to see what would happen to the freed hour - they decided in advance. Thirty minutes of the freed time per rep, per day, was assigned to a mandatory outbound follow-up call block. The other thirty minutes was assigned to absorb a 15 percent year-over-year increase in inbound quote volume.
The first allocation produced revenue that could be traced call by call. At four calls an hour, thirty minutes yields two extra calls per rep per day, or 352 extra calls a month across the team. At a historical 10 percent conversion rate for a live follow-up call, that is 35 additional closed deals a month; at an average order value of $420, that is $14,700 in monthly revenue, tagged in the CRM to the new call block rather than estimated after the fact.
The second allocation produced a cost avoided rather than a revenue gained. The remaining capacity - 88 hours a month across the team - absorbed the volume growth that had been budgeted to require a new junior hire at a loaded $2,300 a month. Because the CFO formally removed that line from the Q3 headcount plan, Tebaa counts it: not because the capacity theoretically could have covered a hire, but because a hire was budgeted and then struck from the plan as a direct result. It is the kind of effect a capacity ledger is built to catch precisely because it is common: the ILO's refined global index of occupational GenAI exposure finds that "one in four workers are in an occupation with some GenAI exposure," which is exactly the scale of quiet capacity reallocation a vanity hours-saved figure is not built to detect.
Add the two together and the real, P&L-visible value of the rollout is $17,000 a month - about seven times the $2,464 vanity figure, and unlike that figure, a number that appears in an actual financial statement.
A Framework, Not a One-Off Calculation
Tebaa generalizes the distributor case into what he calls a capacity attribution ledger, a discipline he applies to any AI rollout regardless of industry. The core moves: time the task before AI with a stopwatch on real work rather than a survey; time it the same way after; decide in writing, before rollout, whether freed capacity goes toward a revenue activity or an absorbed-cost activity; give that destination its own metric starting on day one; count a dollar only once it lands, as booked revenue or a formally removed cost line; and re-check the ledger monthly, since freed time tends to quietly evaporate into busier days if nobody enforces where it goes.
What distinguishes his approach from typical "AI ROI" reporting is the insistence that the allocation decision is managerial, not automatic. The tool creates the possibility of freed time; a specific person, deciding in writing before the tool goes live, is what converts that possibility into revenue or cost savings. Tebaa is explicit that the software gets credit for the hour, and management gets credit for the value.
Reframing the Pilot Conversation
The practical effect on how Tebaa runs AI pilots is that the ledger starts in week one, not at the annual review. Teams working through this kind of measurement discipline alongside a rollout can find implementation-focused resources like Webspot useful, though the framework itself originates with Tebaa's own advisory work rather than any particular vendor. For more of his writing on measuring AI initiatives, see his blog at jonahtebaa.com.
His closing argument is straightforward: the question that matters in an AI review is not how many hours were saved. It is which of two ledgers - revenue moved, or cost avoided - that time landed in, and who made that decision. Without an answer, he argues, the hours are not value. They are simply unclaimed capacity.
