AI adoption strategy: why usage stalls after rollout, and what moves it

Why your AI usage numbers are low

The gap between buying AI and using it opens in the same place in nearly every organization. Leadership adopts early, sees the value, and assumes the rest of the company will follow. BCG's 2025 survey of more than 10,600 people called the result a silicon ceiling: more than three quarters of leaders and managers used generative AI several times a week, while regular use among frontline employees stalled at 51%.

The people who have not adopted are rarely confused about the tools. Most have never heard a plan, many are not sure using AI is welcome, and a growing share are worried about what it means for their jobs. Those are leadership problems, and a license does not touch them.

The manager is the variable you control

Gallup's State of the Global Workplace 2026 found that, inside US organizations investing in AI, employees who strongly agree their manager actively supports their team's use of AI are 8.7 times as likely to say AI has transformed how work gets done in their organization. In an earlier wave of more than 19,000 employees, manager support more than doubled the likelihood of using AI a few times a week or more, and only 28% of employees strongly agreed their manager provided it.

BCG found the same thing from a different angle: the share of employees who feel positive about generative AI rose from 15% to 55% with strong leadership support.

In most organizations that variable is entirely unmanaged, and it is the one you personally control.

Judgment Day, chapter 3

Active support is concrete. A manager who uses the tools where the team can see it, shares what worked and what did not, asks in the weekly meeting what people tried, and protects time for experiments is supporting adoption. A manager who forwards the launch email is not.

Why training alone can make it worse

The usual answer to low adoption is more training, and on its own it can backfire. Gallup's 2026 analysis found that workers who use AI daily or several times a week are more than twice as likely as occasional users to expect their job to be eliminated within five years. People who learn what the tools can do also learn what that might mean for them.

Gallup also found what weakens that link: well-structured management, and the sense that the organization cares about the people doing the work. Teach the tool without addressing the implication and you produce a workforce that is more skilled and more frightened than when you started. Pair every session with an honest conversation about what the organization intends to do with the capability people are building.

Hidden AI use is a trust signal, not just a policy breach

Low official usage often hides real usage. Microsoft's 2024 Work Trend Index found that 78% of people using AI at work brought their own tools. Slack's 2024 survey of more than 17,000 desk workers found 48% uncomfortable admitting AI use to their manager, most often because it felt like cheating or might make them look less competent or lazy. Workers comfortable discussing AI with their manager were 67% more likely to have used it.

A ban drives that use further out of sight. Treat it as information instead. If two people are using the same tool, you have a test group. If twenty are, you may already have an accidental company standard. Set clear rules on which data can go into which tools, then bring the use into the open where people can learn from each other.

When usage is hidden, people assume they are the only one struggling. When it is visible, they feel part of a movement.

Judgment Day, chapter 8

What managers can say when they cannot promise jobs

Adoption asks people to get good at something that may change their role, and most managers cannot promise that the role will survive. Reassurance a manager cannot back loses the room the first time reality contradicts it. What a manager can do is say what they know, what they do not know, and when they will know more, and then keep three promises: people hear about changes from them first, real budget goes to their skills while they are still here, and if a role goes away they work the internal market for that person. The longer version is on our page about AI change management.

A 90-day AI adoption plan

1. Audit

   Count what is actually in use, sanctioned or not, and ask people directly what they believe the tools mean for their jobs. Aggregate the answers and never let an individual response reach a manager. Find out which managers already back the change and which are waiting it out.

2. Arm the managers

   Start with them, before the next round of employee training. Give them language for the job conversation, a short list of uses to try with their teams, and permission to say what they do not know.

3. Experiment

   Pick a few real tasks per team and test AI against a metric that matters, on a fixed clock. Decide in advance what you expect to learn, and write down what happened.

4. Debrief

   What did we learn, what will we keep, and what will we stop? Scale what worked, change what nearly worked, retire the rest, and start the next 90 days from there.

The AI readiness assessment is a free starting point for the audit, and the operating method behind the plan is on one page.

Measure adoption, usage and outcomes every week

Put three numbers on each team's scoreboard and review them in the regular weekly meeting rather than a separate AI review:

• Adoption. The share of the team that has tried AI on real work.
• Usage. How often and how deeply they use it. Intermittent use usually means friction or a lack of trust.
• Outcomes. Whether the results the rollout was meant to move have moved.

Use the numbers to trigger questions, never punishment. When a team's usage falls, the fastest way to find out why is to ask the team in the meeting, and that conversation is adoption work in itself.