Table of contents
Another week, another study finding little to no measurable financial impact from AI.
This one via Laura Bratton at The Information, covering the Wharton School / GBK Collective study of 1,400 employees at large U.S. firms, led by Jeremy Korst and Brian Smith.
The finding: 40% of managers measure AI by hours saved, 27% by revenue, and the 28% reporting strongly positive returns are the ones measuring growth outcomes. Korst calls today’s measurement “more about vibes.” McKinsey & Company said the same in August: record spending, and 94% of enterprises can’t move the earnings needle.
I don’t doubt the data.
And I see positive outliers: large, measured improvement from AI when it’s implemented the right way.
One general observation, one specific one:
General: Measure outcomes (better, faster, cheaper, and/or more of the organization’s actual goals). Activity, such as hours saved, is an input. Logic models turn “we adopted AI” into a number everyone can understand and drive toward.
Dirk Wakeham made exactly this point, better than I did: “Your residents don’t care that you deployed AI. They care that the sink got fixed.” His two questions for every AI deployment are what work went away for the onsite team, and what got better for the person living in the building. Those are the outcome measures. “Hours saved” answers neither.
Specific: AI raises what one person can produce more than 20X. Unless that extra output is shaped and fed into cross-team alignment, that individual productivity goes nowhere. It just means people working incredibly hard without seeing the progress.
