Every client conversation about AI adoption starts the same way. The company bought the licenses. The team got access. Six months later, usage is flat and nobody can explain why the investment isn't showing up anywhere.
Gallup's newest research on managers and productivity finally put numbers on a pattern I've watched for years training teams: adoption doesn't spread up from the frontline. It comes down from the manager a team reports to, and it stalls out the moment that manager stops showing up.
Employees are twice as likely to use AI when their manager does
Employees whose managers actively support AI use are almost twice as likely to use it frequently: 79% versus 46%. Those same employees are 7.4 times more likely to say AI gives them room to do better work, and 8.7 times more likely to say it's changed how work gets done.
Only 21% of employees say their manager actively supports the team's use of AI.
That's a gap. Whether anyone adopts AI depends on whether the people above them are personally using it, not on whether it's been announced.
McKinsey found the same pattern: ownership beats announcements
McKinsey's State of AI research points at the same thing from a different angle. Their AI "high performers," the small group of companies seeing real bottom-line impact, are three times more likely to say senior leaders demonstrate genuine ownership of AI initiatives. High performers are also nearly three times more likely to have redesigned actual workflows around AI, and that redesign is one of the strongest predictors of value McKinsey found.
A memo announcing a new AI initiative doesn't move anything. A leader who changes how they personally work does.
Why employees still tell me AI feels like cheating
Here's something I hear constantly that never shows up in the survey data: employees telling me that using AI still feels like cheating. Like they're getting away with something, or like the output doesn't really count as their work.
I think this is a direct symptom of the gap Gallup found. If nobody above you ever uses AI where you can see it, the tool stays something you sneak in rather than something the work now runs on. Silence from a manager doesn't read as neutral. It reads as permission withheld.
That's the fastest fix available to any manager, and it costs nothing. Use AI on a real task, out loud, in front of the team. Show the messy first draft, not just the polished result. Better yet, show the time it gave you a genuinely bad output or steered you wrong, and walk through how you caught it and fixed it. That does more to normalize AI than any success story could. It tells people the tool is fallible, that checking its output is still their job, and getting it wrong sometimes is just part of using it well. The moment people see someone above them do this openly, "cheating" turns into "how we work now."
Three things managers do differently when adoption actually works
The managers whose teams actually adopt AI aren't doing anything complicated. They're doing three specific things, consistently, where their team can see it:
- Use AI on real work, openly, so the team sees it happen.
- Point to one workflow you've personally changed because of it.
- Treat it as core to how the business runs, not an initiative with a launch date and an end date.
This doesn't rule out training. Nearly every client tells me the same thing: training that's specific to someone's actual job, not a generic tour of the tools, is what makes AI stick in the first place. But even the best job-specific training fades fast if nobody above the room keeps it alive afterward. A manager who references it, uses it, and asks about it later is the difference between a skill the team actually uses and one that gets forgotten by Friday.
None of this requires a bigger budget. It requires managers willing to go first.
