Gallup's own research says the tool was never the deciding variable. The person a worker reports to is.
Gallup surveyed more than 43,000 U.S. employees in the first half of 2026 and found that manager support for AI use is the single factor most strongly tied to employee engagement around it, ahead of having a clear company strategy or using AI frequently. Employees whose managers actively coach AI use are up to 8.8 times more likely to say it helps them do their best work, yet only 28% of managers are currently doing this. The fix isn’t a bigger license count or a longer training deck. It’s giving first-line managers an explicit mandate, real time, and the coaching skill to reinforce AI use inside the work their team already does.
A regional healthcare network turned on an AI charting assistant for eight hundred nurses this spring. The company-wide email went out on a Monday. IT confirmed every license was provisioned by Thursday. A recorded fifteen-minute training video sat on the intranet, and a helpdesk queue stood ready for anyone with questions.
Six weeks later, the adoption dashboard told a different story. Fewer than one in five nurses had opened the tool more than twice. The ones who had were disproportionately the same nurses who’d already been experimenting with AI on their own phones before the rollout even started. One nurse manager later admitted she’d forwarded the training link, moved on to the next item on her agenda, and only found out that afternoon, secondhand, that three of her nurses weren’t sure they were even allowed to paste patient notes into it. Nobody had told her the answer either.
Nobody on that project thought the tool itself was the problem. It worked. It did what the vendor said it would do. And that’s usually where the postmortem goes wrong, because the instinct is to check the technology again: is the interface confusing, is the output good enough, does it need a better model underneath. Those questions rarely find the actual answer.
Ask the people who never opened it why, and the technology barely comes up. Gallup looked directly at why employees don’t use AI at work, and access turned out to be rarely the constraint: among employees who don’t use AI in their role, just 16% point to lack of access, while nearly three times as many, 44%, say they simply don’t believe AI can help with the specific work they actually do.[1]
That’s a belief problem, not a technology problem. And a belief about whether a tool matters to your actual job rarely changes because of a company-wide email. It changes, or it doesn’t, because of what your own manager says and does about it.
This is the part of the AI adoption conversation that tends to get skipped, because it’s more comfortable to audit software than to talk about management. The pattern in the data says otherwise. What determines whether AI adoption sticks has less to do with which tool a company bought and much more to do with what happens between a worker and their manager in the weeks after it shows up.
Gallup has been tracking U.S. workers’ AI use every quarter for a few years now, and the shape of what predicts adoption has become fairly consistent by this point. It isn’t mainly about the tool, and it isn’t really about a worker’s general attitude toward AI either.
In the first half of 2026, Gallup surveyed 43,262 employed U.S. adults about how they use AI at work, whether their organization has adopted it, and what happens around them when they do. A little over half of those respondents, 21,724, worked somewhere that had actually integrated AI tools into its practices.[2] That’s not a small or a casual sample, and what it found doesn’t flatter the way most companies are currently running their rollouts.
Employee engagement rate by AI-support condition, first-half-2026 Gallup workforce survey, N=43,262 (21,724 in AI-adopting organizations). Source: Gallup, “Employee Engagement Remains Flat as AI Adoption Accelerates,” 2026. [2]
The single factor most strongly tied to whether an employee was actually engaged at work wasn’t whether their organization had a clear AI strategy, and it wasn’t even how often they personally used AI. It was whether their manager actively supported the team’s use of it, which put engagement at 48% versus 30% for employees who didn’t say that, an 18-point gap. When all three conditions lined up at once, a clear organizational plan, frequent personal AI use, and active manager support, engagement climbed higher still, to 53%.[2]
Notice what’s doing the work in that combination. A clear plan from leadership matters. Using the tool regularly matters. But manager support is the one variable that shows up as the strongest single predictor on its own, and it’s also the one most rollouts spend the least time and budget on. Most AI rollout money goes toward licenses, a platform migration, maybe a training vendor for a one-off session. Very little of it goes toward giving a first-line manager the time or the explicit mandate to actually coach their team through it.
Picture two teams inside the same company, both handed access to the same AI tool in the same week. On one team, the department head runs a single lunch-and-learn demo, gets a round of applause, and moves on to the next item on the roadmap. On the other, the first-line manager brings one real example into her Monday team huddle every week for a month: here’s something I used it for, here’s where it got something wrong, here’s what I changed before sending it. She asks what got in the way, not just whether people logged in.
A quarter later, one of those teams has quietly stopped using the tool except for the two people who liked it from day one. The other has built it into how they actually work.
Gallup’s numbers back up which of those two scenes is closer to what actually happens across a large sample of workers, not just a plausible story. Within organizations that are investing in AI, employees who strongly agree their manager actively supports the team’s use of it are meaningfully more likely to adopt it in ways that stick.
How much more likely employees are to report each outcome, compared with employees whose managers don’t actively support AI use, among workers at AI-investing organizations. Source: Gallup, “Manager Support Drives Employee AI Adoption,” 2025. [1]
Here’s the uncomfortable part. Only 28% of employees at organizations that have started implementing AI strongly agree their manager actively supports the team’s use of it.[1] Most managers aren’t doing the thing the data says matters most, not necessarily because they don’t care, but because nobody told them it was part of their job, or gave them the time to do it.
A company-wide announcement can put a tool in front of a thousand people. What a worker’s own manager does with it in the weeks after decides whether any of them keep using it.
“Manager” also isn’t one job. A first-line manager running a weekly huddle, a department head setting priorities across six teams, and an exec sponsor who signed off on the budget are three different people doing three different things, and the data above points at only one of them as the daily lever.
| Role | What this role actually changes | The one thing to do differently this month |
|---|---|---|
| First-line manager | Has the standing weekly contact that Gallup’s data ties to the engagement and usage jump. Translates a policy into a specific expectation for one team. | Bring one real AI example into a huddle or 1:1 every week, and ask what got in the way, not just whether the tool was used. |
| Department head | Sets whether first-line managers have the time, priority, and permission to coach on this at all. | Put “coach your team on AI use” on managers’ own goals, not just “confirm your team completed the training.” |
| Exec sponsor | Sets the clear organizational plan Gallup found raises engagement, but doesn’t have the daily contact that turns a plan into behavior. | Fund and protect manager coaching time explicitly, and stop measuring the rollout by license counts alone. |
Three different roles inside “management,” and what each one actually controls in the adoption data above.
This is really the reinforcement stage of adoption showing up in the numbers. Training builds someone’s capability to use a tool. A real workflow gives that capability somewhere to land. Reinforcement, delivered mostly by a manager who notices, asks, and coaches, is what turns capability into behavior that survives past the second week. Adoption is what you see once all three have actually happened. It isn’t something a company can announce into existence.
Ask a CFO to justify an AI rollout budget and the line items are predictable: license seats, maybe a platform migration, sometimes a training vendor for a half-day session. None of that is wrong, exactly. It’s aimed at the wrong stage of the problem.
Gallup’s broader adoption data shows the gap in who actually benefits once a tool is in place. Among employees who use AI, those in leadership roles are far more likely to report an extremely positive effect on their productivity, 21%, than individual contributors are, at 13%.[3] Leaders get more out of the same tool partly because they have more exposure and clearer use cases in their own work, but also because nobody is coaching most individual contributors the way a manager might, in effect, coach herself.
It helps to be specific about what each part of a rollout is actually built to do, because companies routinely pay for one and expect it to deliver the results of all three.
| What it is | What it does | What it can’t do alone |
|---|---|---|
| The tool itself | Gives someone the means to try something new. | Doesn’t tell anyone when to use it, or why it matters for their specific job. |
| Structured training | Builds the underlying capability and teaches what the tool can actually do. | Tends to wear off within weeks if nothing in the day-to-day workflow reinforces it. |
| Manager reinforcement | Turns capability into a habit, through weekly attention, real examples, and coaching. | Needs training and a real workflow to reinforce. It isn’t a substitute for either one. |
The three inputs a rollout budget usually funds unevenly, and what each one is actually responsible for.
None of this means the license or the training budget was wasted money. A worker with no access to the tool and no idea how it works isn’t going to adopt it no matter how good their manager is. What the pattern above actually shows is narrower and more useful: access and training get someone to the starting line. Whether they keep running depends on what happens around them afterward, and that’s a coaching problem, not a procurement one. It’s also why Future Factors builds manager coaching into its Corporate Workshops rather than running a one-off, all-staff session and calling the rollout finished. The manager still in the room three weeks later is the one who decides whether any of it survives.
Training builds the capability. Whether it turns into a habit depends on what a manager does with it the following Monday, not on the training itself.
None of this means throw out the rollout plan you already have. It means checking whether that plan puts any real weight on the one lever the numbers above say matters most.
If two or more of these are true, the rollout is currently aimed at access, not adoption.
Measuring the rollout matters as much as running it, and login counts alone will tell you less than you think. A more honest lens tracks three separate, and separately checkable, questions.
| Stage | What to actually check | A red flag |
|---|---|---|
| Use | Did the team log in and try it at all in the first month? | Usage spikes right after the announcement, then disappears by week three. |
| Persistence | Are they still using it a month later, without anyone reminding them? | Usage only reappears when a manager nudges or a deadline forces it. |
| Impact | Is the actual work different, faster, or better because of it? | Usage looks steady, but nobody on the team can name a task that’s actually changed. |
Logins and active-user counts only answer the first question. Adoption is what the second and third questions describe.
If AI use only comes up at the company town hall, it isn’t reinforced anywhere that actually changes behavior: a recurring 1:1 or team huddle.
This is also where the Tool x Workflows x Behavior idea we use elsewhere on this site earns its place, rather than being bolted on for the sake of mentioning it. A manager who coaches is largely the mechanism that turns a Tool into an actual change in someone’s Workflows and Behavior, not a fourth ingredient added on top. If you want the individual-level version of that framework in full, we’ve laid it out separately in what makes someone an AI-powered professional.
If you’re at the point of actually building the rollout itself, the practical checklist for what belongs in an enablement kit, and why a champions network tends to outlast a single training session, is worth reading alongside this piece. We covered that ground separately in our AI user enablement guide, which stays deliberately tactical where this piece stays deliberately about the psychology behind why any of it works. For the wider pattern of why enterprise AI investment and enterprise AI results keep drifting apart, our companion piece, what the 2026 numbers actually reveal, covers that ground in full. And if you’re a manager wondering what this looks like from your own seat rather than the program level, we’ve written separately about how to use AI to be a better manager.
None of this needs to wait for a full program relaunch. It’s a smaller ask, and it’s closer to what actually moves the number above:
Based on Gallup’s own research, yes. Most employees who don’t use AI at work aren’t blocked by access: only 16% say that’s the reason. Most say they don’t believe the tool applies to their actual work, which is a belief someone has to change, and a company-wide announcement rarely does that on its own.
A large one. Gallup found that employees whose managers actively support AI use are 2.1 times as likely to use it a few times a week or more, 6.5 times as likely to say the tools are actually useful, and 8.8 times as likely to say AI helps them do their best work, compared with employees whose managers don’t. Only 28% of managers are currently doing this.
Spending the rollout budget on licenses and a training session while leaving first-line managers with no explicit mandate, no extra time, and no coaching of their own. The tool and the training get someone to the starting line. What their manager does afterward decides whether they keep running.
The manager-support pattern shows up broadly, but the size of the effect varies by role. Gallup found leaders and people in technical or professional roles report bigger productivity gains from AI than employees in service or administrative roles do, which suggests a manager’s coaching matters even more where the use case isn’t obvious on its own.
A manager who coaches is largely what turns a Tool into an actual change in someone’s Workflows and Behavior, rather than a fourth ingredient on top of the other three. Access to a tool alone rarely moves any of that by itself; a manager’s attention in the weeks after rollout is one of the more reliable ways it actually happens.
The manager-effect statistics in this piece are drawn directly from Gallup’s own published workplace research, fetched and checked against Gallup’s own site rather than a secondary summary: “Manager Support Drives Employee AI Adoption” (November 2025), “Employee Engagement Remains Flat as AI Adoption Accelerates” (July 2026, updated August 2026), and the role-based productivity breakdown in “Rising AI Adoption Spurs Workforce Changes” (April 2026). The title and angle were inspired by a Forbes column covering related Gallup research; this piece independently verified its own figures directly against Gallup’s site and builds its own analysis, framework, and recommendations rather than summarizing the Forbes piece or reusing its framing. The manager behavior ladder and the tool/training/reinforcement breakdown are Future Factors’ own framework, not a finding from Gallup.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.