How to Get Employees to Use AI: A CEO's Playbook
To get employees to use AI, stop treating it as a software rollout and start treating it as a change to specific work. Pick a few recurring workflows, rebuild them with AI inside, have managers use it in the open, and measure use inside those workflows instead of counting logins. Access alone does not create usage.
That last point is now well documented. In a February 2026 Gallup survey of 23,717 US workers, about half used AI once a year or not at all, and only about 3 in 10 were frequent users. Among non-users, 46% said they simply prefer to keep working the way they do now. That is not a skills gap. It is a habit, and habits change when the work changes.
Why don't employees use the AI tools you paid for?
Most companies buy licenses, run a lunch-and-learn and wait. Then usage spikes for two weeks and fades. The reasons are predictable, and most of them are not about the tool.
- No reason to change. In the same Gallup survey, about a quarter of non-users had tried AI and did not find it helpful, and only about 2 in 10 said they didn't feel prepared. People who don't see AI in their actual work go back to the old way.
- No clear permission. BCG's AI at Work 2025 survey of 10,600 workers found only 25% of frontline employees say leaders give enough AI guidance, and 54% would use AI tools even without authorization. When the rules are vague, the careful people hold back and the bold ones go around you.
- A quiet social cost. Four preregistered experiments published in PNAS in 2025 found that people who use AI at work are judged as lazier and less competent than people who get the same help from a colleague. Nobody says this out loud in your company, but your team feels it.
- Tools that don't fit the work. MIT's NANDA initiative reported that about 95% of organizations saw no measurable return from generative AI, and pointed to brittle workflows and tools that don't fit daily operations as a core cause.
A useful definition before going further: AI adoption means people using AI as a normal part of recurring work, not trying it once. Everything below is aimed at that.
Is AI adoption a training problem?
Partly, and less than most CEOs assume. Training matters: BCG found that employees with five or more hours of training, especially in person with coaching, were significantly more likely to become regular users. But only 36% of BCG's respondents felt adequately trained, and Gallup's non-users rarely named preparation as their main reason.
The bigger levers are managers and workflow fit. In Gallup's 2026 workplace data, 78% of employees who strongly agree their manager actively supports AI use are frequent users, against 44% of everyone else. Only 36% strongly agree their manager does this. Gallup reports these as associations, not proof of cause, but the gap is too large to ignore. Training teaches people how. Managers and redesigned workflows give them a reason.
What does the CEO have to do personally?
Here is a connection most adoption guides miss. Gallup found that 51% of leaders were frequent AI users by Q2 2026, against 26% of individual contributors. The PNAS researchers found that the social penalty for using AI shrinks when the person judging is a frequent AI user, and it disappears for digital tasks that clearly suit AI. Put those together and your role is clear: your team won't feel safe using AI until the people who judge their work are visibly using it too.
That makes your own use a management tool, not a private productivity habit. In practice:
- Say out loud what you used AI for this week, in a leadership meeting or a short note. Be specific about the task and what you still checked yourself.
- Ask "did you try AI on the first draft?" as a normal question, the way you would ask whether someone checked last quarter's numbers.
- Credit AI-assisted work in reviews. If good work done faster with AI gets less credit than slow work done by hand, people will hide their use or stop.
- Publish one page of rules: which tools are approved, what data can go in, and what needs a human check. Short rules beat a long policy nobody reads.
How do you get a team to actually use AI? A six-step workflow
This is how a company moves from scattered experiments to AI that is part of the job. It works on one workflow at a time, which is why it sticks.
- Pick three recurring workflows, not three tools. Look for work that happens at least weekly, has a clear owner, and produces output someone can check quickly. Good candidates are first-draft proposals, the weekly pipeline review and month-end variance commentary. Gallup found the highest share of positive productivity reports in coding (77%), slide decks (76%) and analytics (75%), which is a good hint about where to start.
- Map the current workflow step by step. Spend 30 minutes with the people who do the work. Write down each step, who does it, which systems it touches and roughly how long it takes. You need this baseline to know later whether anything improved.
- Redesign it with AI inside. Decide which steps AI drafts, which steps a person checks, and where the data comes from. Put the AI in the tool people already work in. Then retire the old way: if the old template still exists, people will use it.
- Name one owner per workflow. Not IT and not a committee. Someone who understands the business problem and can work with IT on the setup. This is the product role most companies are missing, which we covered in Should IT Own AI?
- Train inside the workflow, with peers. Teach people on their real work, not generic prompting. ServiceNow, training 28,000 employees, recruited 1,000 avid AI users to teach their colleagues. At 50 or 300 people, the same idea is two or three volunteers per team.
- Measure use in the workflow every week for eight weeks. Review at weeks two, four and eight. Fix what blocks people, and only roll the pattern to the next workflow once this one holds.
What should you measure to know adoption is working?
Most AI dashboards count activity. You want to count changed work. Here is the swap:
- Instead of licenses assigned: track the share of runs of the target workflow that used AI.
- Instead of weekly logins: track the cycle time of that workflow, against your baseline.
- Instead of prompts sent: track the rework or error rate on the output.
- Instead of training completion: track the number of distinct tasks each person uses AI for.
The last item matters more than it looks. Gallup found that among employees who use AI for one or two purposes, 45% report a positive effect on productivity. Among those using it for seven or more, it is 90%. Breadth of use and perceived value rise together, so a practical goal is moving each person from one or two uses to three or more.
How does adoption change as your company matures?
We describe AI maturity in four stages: Scattered, Experimenting, Operating and Compounding. Adoption looks different in each, and knowing where you are tells you what to fix.
- Scattered: individuals use personal tools on their own. MIT's NANDA research found workers at over 90% of companies use personal AI tools for work, while only 40% of companies had bought an official subscription. Usage exists, but you can't see it or build on it.
- Experimenting: licenses and pilots are in place. Usage is spiky and depends on a few enthusiasts. This is where most companies stall.
- Operating: a handful of workflows have been redesigned, each with an owner and a weekly number. Use no longer depends on enthusiasm.
- Compounding: each redesigned workflow makes the next one faster, because the data, rules and people who know how are already in place.
The six-step workflow above is how you get from Experimenting to Operating. More licenses won't get you there. Redesigned work will.
What mistakes stall AI adoption?
- Mandating use without changing the work. A rule that says "use AI" with the same templates, deadlines and approvals produces box-ticking, not better work.
- Banning tools instead of approving good ones. With more than half of workers willing to use AI without permission, a ban mostly pushes use out of sight.
- Rolling out to everyone at once. Three workflows done well teach you more than thirty done halfway. Your size helps here: MIT's research found mid-market firms took about 90 days to go from pilot to full rollout, against nine months or more for large enterprises.
- Treating adoption as IT's job. IT can make the tool safe and connected. It can't decide how a sales manager runs a pipeline review. That decision belongs to the business, with you backing it.
The short version
AI that isn't used isn't AI. It is a line item. Your employees will use AI when it sits inside work they already do, when their manager uses it in the open, and when someone owns the result. Start with three workflows, measure them weekly, and let the results do the persuading.
If you want an outside view of which workflows in your business to start with, Journey's free CEO AI audit is a good place to begin.
Frequently Asked Questions
Mostly because their work hasn't changed. In Gallup's February 2026 survey, 46% of non-users said they prefer to keep working the way they do now, and only about 2 in 10 said they felt unprepared. Licenses add an option. People adopt AI when it sits inside a workflow they already run, their manager uses it openly, and the old way is retired.
No. Training helps: BCG found employees with five or more hours of training, especially with in-person coaching, were much more likely to become regular users. But training teaches how, not why. Manager support and workflow fit matter more. Gallup found 78% of employees with strongly supportive managers use AI frequently, against 44% of everyone else.
Use AI visibly and talk about it. Research in PNAS found people who use AI are judged as less competent, but the penalty shrinks when the evaluator uses AI too. So share what you used AI for, ask whether teams tried AI on first drafts, credit AI-assisted work in reviews, and publish a one-page set of rules on approved tools and data.
Measure changed work, not activity. Instead of licenses or logins, track the share of a target workflow's runs that used AI, the workflow's cycle time against a baseline, rework or error rates, and how many distinct tasks each person uses AI for. Gallup found people using AI for seven or more purposes were twice as likely to report productivity gains.
Plan on about eight weeks per workflow, with reviews at weeks two, four and eight, before rolling the pattern to the next one. Smaller companies have a speed advantage: MIT's NANDA research found mid-market firms took about 90 days to go from pilot to full rollout, against nine months or more for large enterprises.
