If you run a bank, an insurer, a government agency or a large firm, you may receive a report on AI use every month. It has a line that goes up. You're asked to read that line as progress.
I want to be fair to these reports. They answer a real question: is the money spent on AI being used? But they answer it with counts of use. A count of use does not show the thing that matters most to a leader, which is whether people decide well with what the AI gives them.
What the report counts
Reports on AI use often count the same kinds of thing: licences issued, people active in the month, prompts sent or tasks completed, and hours said to be saved.
All of these count use. They tell you that people opened the tool and that something came out of it. They're worth having. A licence nobody opens is money you can save, and a team that never starts may need help to start.
Use is where the work begins. The decision comes later, and a person makes it.
What it leaves out
Between an AI answer and the action that follows, a person decides. That part doesn't appear in the report.
The report doesn't show what was checked before an AI answer was used, or what was changed or rejected. Nor does it show who decided to act on the answer, or what happened after.
A rising line of use says nothing about any of these. It can rise when people check every answer with care. It can rise in the same way when people paste answers into their work without reading them. From the report, the two look the same.
A credit team that drafts every client summary with AI will push the usage line up, whether or not anyone opens the figures behind each summary.
Why that matters
Two studies bear on this.
KPMG and the University of Melbourne surveyed more than 48,000 people in 47 countries between November 2024 and January 2025. Of the employees in that study who use AI at work, 42% said they relied on AI output without evaluating its accuracy sometimes or more often. A report on use counts those people in the same way as people who check every answer.
The second study is about results. In a 2024 meta-analysis of 106 experiments, Vaccaro, Almaatouq and Malone found that people working with AI did better than people working alone. But a person and an AI working together did worse than the better of the two working alone, most of all on decision tasks.
That study doesn't say AI makes decisions worse. Across the experiments, people with AI did better than people without it. My own reading is this: more use of AI is not the same as better decisions. Adding AI to the work leaves a decision for the person to make, about when to rely on the answer and when to set it aside. A report on use can't see that decision.
Do you know how your people are deciding with AI?
What I'd ask for beside the report
Keep the report. Ask for something beside it, and ask for it in a meeting rather than a dashboard.
Start by asking the team to describe one recent decision made with AI, step by step, from the answer it gave to what was done. Pick a decision that mattered to the business. The description will show whether anyone looked at the basis of the answer before acting on it.
A second question is about the answers that were changed or rejected. Ask which AI answers were changed or rejected last month, and why. If the honest answer is none, ask how anyone would have noticed a wrong answer.
The last question is about ownership. Ask who owns the result when an AI-assisted decision goes wrong. If the answer names the tool or the data, no person has yet taken responsibility for that decision.
None of these needs a new system. They need a leader who wants to hear the answer, and a team that knows it's safe to give it.
You may decide you want a regular report on how people decide, not just on how much they use AI. That's a measurement question. It should be answered by a published method, not by me. AIR APAC, the institution I founded, publishes its methods openly. I don't measure or score anyone's decisions.
What I won't claim
I'm not saying reports on AI use are useless or that you should stop reading them, only that they leave out how people decide.
I give keynotes, briefings and workshops on this.