Your managers are the bottleneck. Nobody is training them.
Pick any AI training catalogue. You will find courses on prompt engineering. Courses on Copilot. Courses on “AI fundamentals” that explain large language models in twelve minutes. You will find almost nothing on the actual job of managing a team through AI adoption.
This is the gap. And it is the gap that determines whether your AI rollout works or quietly fails.
Your managers are the people deciding which tasks get automated, which staff get reassigned, which performance conversations need to happen when output quality drops or jumps. They are the people fielding the “is my job safe?” questions over coffee. They are the people who will either build psychological safety around AI use or destroy it by accident.
And almost none of them have been trained for any of this.
This post is for the L&D, P&C, or HR lead who has rolled out AI tools to their organisation and is now realising that the rollout was the easy part.
Why prompt training is not the answer
The instinct, when senior leadership says “we need to train our managers on AI,” is to commission a prompt engineering course. This is the wrong instinct.
Prompt engineering is a fifteen-minute skill. Anyone with a working browser and ten minutes of curiosity can pick it up. It is not what managers are struggling with.
What managers are struggling with is everything around the prompts:
- A team member submits a report that is clearly AI-drafted. Is that fine? When is it not fine? How do you have that conversation?
- Half the team is using AI to triple their output. The other half refuses to touch it. How do you do performance reviews fairly?
- A senior team member is anxious that AI will make their role redundant. How do you respond honestly without either lying or making it worse?
- The org has rolled out Copilot. Adoption is at 12 percent. How do you actually drive uptake without it feeling forced?
- You suspect a team member is using AI to fake expertise they do not have. How do you tell?
None of these are prompt problems. All of them are management problems. And they require a different kind of training than the “here is how ChatGPT works” content currently flooding the market.
What a good AI training course for managers should cover
A custom AI course for managers should do five things.
1. Reframe AI as a management challenge, not a technology one
Most managers walk in expecting a technology course. The first job of the training is to shift that frame. AI adoption is not about understanding the technology. It is about leading people through a workflow change that touches their identity, their workload, and their sense of where they stand.
The opening of the course should make this explicit. The managers who succeed with AI are not the ones who understand the technology best. They are the ones who manage the human side best.
2. Cover the new performance conversations
There is a class of performance conversation that did not exist three years ago and that almost no manager has been trained for:
- “Your output has jumped significantly. I want to understand what changed.”
- “I noticed this report was AI-drafted. Walk me through how you used it.”
- “Your peer is producing similar work in half the time. Let’s talk about how they’re doing that.”
- “I need to understand which parts of this analysis are your judgement and which are the model’s.”
These conversations need to happen. Most managers avoid them because they do not have the script. The course should give them the script. Not as a lecture, but as scenarios where they make the call and see what happens.
3. Handle the anxiety, honestly
The single most damaging mistake managers make with AI is reassuring their teams that “nothing will change” when something obviously will. The team knows. The manager knows the team knows. Trust evaporates.
Good training teaches managers how to have the honest version of this conversation. What you can say. What you should not promise. How to acknowledge uncertainty without spiralling. How to be the person who tells the team the truth before someone else does it badly.
This is not a soft skill. It is the difference between a team that experiments openly with AI and a team that uses it in secret because they are afraid of looking replaceable.
4. Drive adoption without faking it
Most AI rollouts fail at the manager layer. The tools are deployed, the training is offered, the dashboards show single-digit adoption, and senior leadership starts asking why.
The answer, almost always, is that managers were not given a role in the rollout. They were sent a Slack message and a calendar invite, and now they are expected to drive behaviour change in their teams with no incentive, no script, and no support.
The training should cover how to do this properly: how to identify the two or three workflows in your team where AI will make an immediate difference, how to run a low-stakes experiment, how to talk about results without overclaiming, how to bring along the people who are resistant. This is where AI training overlaps directly with leadership development, and where standalone prompt courses fall short.
5. Spot the misuse patterns
AI-assisted fabrication is real and growing. Staff using AI to fake expertise in areas they do not actually understand. Reports that look polished but rest on hallucinated data. Code that runs but does the wrong thing in subtle ways.
Managers need to know the signs. Not because they need to become detectives, but because they need to know when to verify and when to trust. The training should walk through real patterns: the over-confident summary, the suspiciously even tone, the absence of the rough edges that human work usually has.
How long should the course be?
Longer than acceptable use. Shorter than you think.
Twenty to thirty minutes is the sweet spot for self-paced content. The scenarios are where the value is. Generic content about “leading through change” can be cut to almost nothing. Most managers have done that training before, in some form. What they have not done is apply it to AI specifically.
If your organisation can support it, the highest-impact version is a blended approach: a self-paced module for the conceptual content, followed by a facilitated workshop for the scenarios. The self-paced piece is what we build. The workshop is something your L&D team or an external facilitator can run.
What it should cost
A custom AI training course for managers, built properly, sits in our Tier 2 bracket: $7,500 plus GST. This is the right tier because the course needs branching scenarios, role-specific examples, and assessment that goes beyond knowledge checks.
If you want simpler treatment, a straightforward awareness course without the scenario depth, Tier 1 at $5,500 is viable. We have built both. The Tier 2 version performs noticeably better in practice because the scenarios are where managers actually change their behaviour.
What you should be wary of: courses sold to you for $15,000 to $25,000 that take three months to produce. By the time they launch, half the AI examples are out of date. Speed matters here more than in almost any other training category.
When to commission it
Ideally, before your AI tools roll out. Realistically, in the first ninety days after.
The window matters. Managers form their AI habits, and their teams’ habits, in the early months of a rollout. If they form those habits without training, you spend the next year unlearning bad patterns. If they form them with training, you get a faster, calmer adoption curve and significantly fewer “we need to revisit how we’re using this” conversations later.
If you have already rolled out AI and you are now seeing the adoption stall, the anxiety, or the misuse patterns described above, commission the training now. It is not too late, but every month makes the unlearning harder.
What to brief your developer on
When you commission an AI course for managers, bring:
- The AI tools your managers’ teams are using. Specific products, specific versions.
- Your organisation’s position on AI use. Encouraging, neutral, restrictive. The training has to match.
- The five hardest situations your managers have actually faced. Anonymised. These become the scenarios.
- Your existing leadership development content. So the new course extends what is already there rather than contradicting it.
- The metric you are trying to move. Adoption rate, sentiment, output quality, retention. Different metrics need different course structures.
That is enough for a single kickoff call. Three weeks later, the course is in your LMS and your managers have something to actually work with.
The bigger picture
Manager training is the second layer of the AI capability stack. The first is AI acceptable use training: get the policy training in place across the whole organisation. The second is this one, focused tightly on the managers who will make or break adoption.
After that comes function-specific training for the teams using AI most heavily, and governance content for compliance and risk. But manager training is the layer most organisations skip, and it is the layer that costs the most when it is missing.
When you are ready, get in touch. Three weeks from kickoff to launch. Fixed price. No drama.