Most organisations launched AI with licences, training videos and a town hall. A year later, use is often patchy. Some teams have changed how they work; others have barely started.
Four large 2026 studies help explain the gap. They come from different organisations, with different methods. Yet they keep landing on the same point: the manager decides whether adoption holds.
What Gallup found about manager support
Gallup's 2026 page on AI and workplace productivity draws on its surveys of US employees. It was first published in July 2026 and updated in November 2026.
In AI-adopting organisations, 78% of employees who strongly agree their manager supports AI use are frequent users. Among everyone else, the figure is 44%.
Those employees are also 9.3 times as likely to strongly agree that AI has transformed how work gets done. And they are 7.8 times as likely to say AI gives them more chances to do what they do best.
Yet only 36% of employees in AI-integrating organisations strongly agree that their manager actively supports AI use. Gallup sells management advice, so it has an interest in this finding. The size of the gap is still hard to ignore.
What Microsoft found about modelling
Microsoft's 2026 Work Trend Index included a separate study of 1,800 workers. Where managers modelled AI use, workers reported AI value 17 points higher and critical thinking 22 points higher. Trust in agentic AI was 30 points higher.
Where managers created psychological safety for experimentation, readiness and value scores were up to 20 points higher. Those workers were also 1.4 times as likely to be high-frequency users of agents.
These are self-reported associations from a company that sells AI tools. They do not prove that modelling causes adoption. They do show what employees notice.
What SHRM found about communication and trust
SHRM surveyed 5,875 US workers in March and April 2026. It asked whether people heard about AI before it arrived in their work.
Prior communication reached 74% of directors and above and 55% of managers. Only 1 in 3 individual contributors said the same.
Trust followed a similar pattern. Overall, 61% trust senior leaders on AI, but only 47% of individual contributors do.
The people furthest from the decisions heard about AI last. Managers are usually the only leaders those people talk to every week.
What Wharton found about training
The Wharton and GBK Collective study surveyed 1,416 people in US enterprises in July 2026. It compared leaders' views with those of entry and mid-career employees for the first time.
78% of leaders said AI training is adequate. Only 56% of employees agreed.
That gap matters for managers. If the organisation believes training is done, the manager is often the only person who can see that it is not.
Build this capability in your organisation
Growth Performance designs bespoke programmes that turn insight into measurable behaviour change.
The four practices that matter
Gallup lists four things effective managers do. They line up closely with the other three studies.
- Support AI use at team level. Say clearly which tasks AI should help with, and which it should not. Silence reads as disapproval.
- Model AI use. Use AI in team meetings and talk about where it helped and where you had to correct it.
- Answer questions. Make time for the basic questions people are embarrassed to ask. Treat them as useful, not as a sign of weakness.
- Connect AI to daily work. Link each use to a real task the person does every week, not to a generic demo.
None of this needs a new budget. It does need managers who are confident enough to go first.
What this looks like in a normal week
None of the four practices needs a programme to start. They fit into the meetings managers already run.
In a weekly team meeting, spend five minutes on one real task where someone used AI. Ask what worked, what needed correcting and whether the rest of the team could use the same approach.
In one-to-ones, ask a simple question: where would AI help you most next week? Then agree one small thing to try before the next conversation.
When someone admits they do not understand a tool, thank them. That response tells the whole team that questions are welcome.
When AI gets something wrong, say so openly and show how you caught it. Checking is part of good use, not a sign of failure.
What senior leaders can do
Managers cannot do this alone. They need to hear about AI changes before their teams do, so they can answer questions with confidence.
They also need permission to spend time on it. If every hour is already allocated to delivery, modelling and coaching will be the first things to go.
How to find out where your managers stand
If you lead managers, start with evidence rather than impressions. The Manager Capability Heatmap maps a group of managers against fifteen behaviours in about ten minutes, so you can see where support is thin.
If you are a manager, the AI Leadership Readiness Scorecard asks twelve questions on how ready you are to lead with AI.
For the wider picture of why adoption stalls, see the Adoption condition. If you want to build these practices with your team, our team events for 2026 and 2027 can be planned around them.
A note on the evidence
All four studies are surveys, and three are US-based. Microsoft's covers ten markets including the UK. None can prove that manager support causes adoption; people who already use AI may also rate their manager more highly.
What makes the finding credible is agreement across sources with different methods and different interests. When Gallup, Microsoft, SHRM and Wharton point the same way, the direction is worth acting on.
Try this now
- In your next team meeting, show one real task where you used AI and what you corrected.
- Ask each person which task they would most like AI help with.
- Map your managers with the Manager Capability Heatmap.
