Most organisations I talk to can now show impressive AI adoption numbers. Licences are rolled out, usage dashboards are green, and almost everyone has tried a chatbot at least once. Yet when I look at what has actually changed in how work gets done, the picture is much thinner.

Recent research published in Harvard Business Review followed how 2,500 employees used large language models over eight months. What separated the strongest users was not seniority or technical skill. They approached AI ambitiously, treated it as a reasoning partner rather than an automation tool, and gave it complex tasks with clear objectives.

That matches what I have seen. Most people use AI to do the same things they always did, a bit faster. A small group uses it to do things they would not have attempted at all, such as testing several versions of a launch plan or stress-testing a forecast from a competitor’s point of view. That is a different category of value.

Why the gap does not close on its own

Tool training teaches people where the buttons are, not how to think with a new kind of colleague. And many leaders still measure what is easy to see, how often a tool is opened, instead of what it changed.

What I would do as a leader

• Find your strongest users and make their work visible.
• Move the metric from usage to outcomes: cycle time, quality, decisions made differently.
• Give people permission and time to try ambitious things with AI.
• Redesign one workflow end to end, instead of adding AI to every workflow a little.

In our own work on GenAI-enabled content and medical, legal and regulatory review, the step change came from redesigning the process around the technology, not from the technology alone.

Do you know who your strongest AI users are, and what are you doing to multiply them?

Source: Hallman, Kowaleski, Puvvada & Schmidt, “What the Best AI Users Do Differently”, Harvard Business Review, March 2026.