The Information reports on what the Wharton-GBK Year Four study found when it asked how companies actually measure the return on their AI spending. Most are tracking whether the technology made people faster — hours saved, errors avoided — rather than whether it produced revenue. Far fewer track money coming in.
How the piece reports what Mindspan Labs founder Jeremy Korst, a co-author of the study, told them about the measurement itself:
Jeremy Korst, one of the Wharton study's authors, said it was "very difficult" to find "actual, materially measurable financial impacts" from AI spending.
He added that measuring returns today is more about managers' "perception" of a return than about hard numbers, and that for a technology this early in its rollout, that is "normal" and in line with previous adoption cycles.
The way we see it, both halves of that matter and they pull in opposite directions. Measurement discipline is a Ring 2 question — it is how an organization decides what counts as working, and it is the muscle that turns AI ambition into results someone is accountable for. An organization that only counts hours saved will keep funding efficiency and keep wondering why the income statement never moves, because it has never once asked the revenue question out loud.
The other half is the correction to the panic. Early returns that show up as perception rather than as a line item are not evidence the technology failed. They are what every prior platform shift looked like at this point. The mistake is not that the number is soft. The mistake is leaving it soft on purpose, because a soft number cannot be argued with and nobody has to own it.
So what: Pick one AI investment and write down, before the next quarter starts, the revenue or cost line it is supposed to move and who answers for it. Keep measuring hours saved. Just stop letting it be the only thing you measure, because it is the number that can never disappoint you.