Managers Are the Bottleneck: AI Training for People Leaders

I’ve run enough AI training workshops to see the pattern clearly: the organisations where adoption stalls aren’t being held back by resistant employees. They’re being held back by managers who haven’t been given what they need.

Managers are the translation layer between company policy and daily practice. They model behaviour, reinforce norms, coach their teams, and - whether they intend to or not - set the ceiling on how far their people will go with any new capability. When a manager is confident and informed about AI, adoption accelerates. When they’re uncertain, it quietly stalls. And here’s the thing that organisations keep missing: most AI training investment goes into individual contributors and almost nothing goes into the people who manage them.

That’s backwards. And it’s costing organisations real value.

The pattern that plays out everywhere

A new AI tool gets announced. Individual contributors are curious - some genuinely excited. But their manager hasn’t had time to try it, isn’t sure what’s allowed, and has three other priorities competing for attention. So they neither encourage nor discourage it, and in that vacuum of guidance, usage stays sporadic.

Some managers go further and actively discourage AI use until they “understand it better” - a moment that never arrives without a deliberate intervention. Meanwhile, the whole team defaults to shallow, generic usage that produces underwhelming results and quietly erodes confidence in the tools.

What leadership-specific training actually looks like

Managers need something different from the training you run for frontline employees. They need to understand AI well enough to have a strategic opinion about it - what it’s useful for in their team’s specific context, what the risks are, what the company’s policies require, and how to evaluate whether it’s being used well.

Practical experience matters here. A manager who has used the tool to draft a performance review, analyse a report, or prepare for a difficult conversation has a fundamentally different relationship with it than one who only watched a demo. That experiential gap is the gap you need to close.

Coaching skills are the missing piece

Once managers have that foundation, they need coaching skills - and this is where most training programmes completely fall short. Reviewing an AI output isn’t the same as reviewing a piece of writing. Coaching an employee to use AI well requires understanding the four-part prompting framework: role, context, standards, goal.

Managers who can reference that framework in a one-on-one conversation - who can look at a weak output and diagnose whether the prompt lacked context or specificity - become genuinely useful coaches rather than vague encouragers. That’s a trainable skill. Invest in teaching it.

Make adoption visible

Give managers simple tools to track AI usage in their teams - frequency, task types, time saved, quality improvements. Review these in team meetings. When adoption is visible, it becomes a topic of conversation. When it’s invisible, it stays optional. A shared spreadsheet that teams update weekly is often enough to shift a team from passive awareness to active practice.

Also give managers permission to experiment. Provide sandbox environments where they can practise before coaching others. Make explicit that first attempts will be imperfect - that’s normal and expected. Creating psychological safety for managers to fail privately before performing publicly is one of the most underrated elements of any training programme.

Peer learning beats top-down training

Cross-functional manager forums accelerate learning in ways that formal sessions rarely do. When a finance manager hears from a sales manager about how they’re using AI for pipeline reviews, the relevance is immediate. Connect managers to your internal ambassador network for ongoing peer support between formal sessions.