Managers set your AI culture. Right now, most are setting the wrong one.
AI Adoption Series - part 3
Welcome to FullStack HR. This is where you come to go beyond the hype and actually learn alongside 10,000+ other smart, curious leaders who are done with surface-level AI talk.by subscribing here:
Happy Tuesday,
Back at the Adoption-series. My initial hypothesis was that this would be a 4-parter, but I realize now that this might be an ongoing thing for much longer…there’s much to be said and lots of learning to be had.
I thought I’d let you in on the process of writing these as well, I “write” them by talking! The three parts that you can now read came from me talking for roughly 3 hours. I’ve then asked Claude here specifically to act as my editor, so you won’t have to suffer through my hour-long conversations. The 3 hours were recorded while driving (safely in my HiDock P1) and while it might seem crazy to talk to yourself like that, I’ve found it to be a great way of thinking.
It’s me through-and-through. Claude has only structured the text (and in all honesty also removed my sometimes quite extensive cursing when I get agitated around why AI adoption is so slow…).
Anyways, thought that might be interesting to know!
Now let’s dive into part 3!
If you haven’t read Part 1 and Part 2 , go back and read those first.
This builds on them.
In Part 1, we covered getting leadership buy-in and structure. In Part 2, we covered the basics of what you need to do and we eluded to manager training in that part. But now we get practical: How do you actually train managers? Because without managers on board, imho, nothing else matters.
Why Managers Are the Critical Layer
Here’s what I’ve seen repeatedly: Organizations train employees, wonder why adoption stays at 10-30%, then realize managers aren’t using AI themselves.
Research backs this up. Having a manager who uses AI is the single biggest predictor of employee adoption. Not the tool you choose. Not how good your training is. Whether your boss uses it.
This is like any organizational initiative. What managers embrace, what managers lead with, that’s what determines success. And look, if you don’t have your leadership team bought in from Part 1, this gets significantly harder. But assuming you have that foundation, managers are your next critical layer.
I used to think we could skip this step. Train everyone at once, let enthusiasm spread. I was wrong. Managers first. Always.
The Two Things Every Manager Needs
Managers need to be competent in two distinct areas. Not one or the other. Both.
First, they need hands-on capability with AI in their own work. How it works for them personally. In their actual daily work. You can’t lead what you don’t understand. A manager who’s never used AI trying to guide employees using it? That’s a blind person leading blind people.
There’s a practical aspect here too. AI can genuinely help with the administrative burden that crushes most managers today. Managers are sitting in the middle of too many stakeholders already. Lots of documentation. Lots of administration. Things get piled on them constantly.
AI can help concretely with documentation that actually matters, prepping for employee conversations, simulating difficult negotiations before they happen, following up on goals across the team, transcribing meetings and pulling action items, leading better meetings with less prep time. I’ve seen managers experiment with all of this. Some with great results. The point is there’s a practical value proposition here specifically for managers, beyond just “you should know how this works.”
Second, they need to know how to lead teams that use AI. This is the part most organizations completely miss. They train managers to use AI. They don’t train managers to lead in a world where their teams use AI. These are completely different skills.
The latter means setting expectations for team members who’ve found 10% efficiency gains, handling the person who’s skeptical or scared, spotting when someone’s using AI too much and removing all personality from their work, managing the gap between AI-native employees and those struggling to adopt, deciding what to do with the time people save, and following up when people use unauthorized tools.
You can’t just assume managers will figure this out. They won’t.
The Training Structure That Works
So how do you do this? Well as I see it, you need minimum 3 hours total for basic capability on the two capabilities mentioned above, working with AI as a manager and leading teams/individuals that are using AI. But three hours is a stretch. Better is at least two sessions of 3 hours each. Even better is fours sessions over two months, a 2-4 hours each.
Sounds like a lot? It is! This requires time. It’s a change management project and thus people need time to change. It doesn’t happen overnight.
What to do in the sessions? (This is the “quick” version.)
The first session focuses on building managers’ own capability. This is about managers becoming competent users themselves. You spend the first hour on foundations, what AI can and can’t do, when to use it. The second hour is hands-on practice with manager-specific use cases. The third hour is managers applying it to their actual upcoming work. Critical point: This must be hands-on, not theoretical. Managers leave with at least 3 specific ways they’ll use AI this week in their actual work.
The second session focuses on leading teams using AI. Requires pre-work either by the lead team and/or HR on that these expectations are clearly defined (see part 1). This is about how to lead when your team uses AI at different levels. First hour covers setting expectations, what does good AI use look like, what’s too much, what’s too little. Second hour addresses handling the gaps between the AI-native employee and the skeptical one. Third hour works through practical scenarios with role-play of difficult conversations. Why this works is you’re giving managers actual tools and frameworks, not just awareness.
What Can You Further Cover?
Shadow AI. The shadow AI usage problem is massive. A KPMG study found that 57% of employees don’t tell their boss they’re using AI. Why? Fear of being replaced, fear of being questioned about their competence, fear their work will be devalued. What this creates is people using ChatGPT when your organization said to use Copilot. They use unauthorized tools. Sensitive data leaks. You have no visibility into what’s actually happening.
I know several organizations that looked at their outgoing web traffic. Top destination? ChatGPT. But they paid for Copilot and told everyone that’s what to use. That tells you something about this question.
Managers need clear guidance on what tools are approved and why, how to have conversations when they discover unauthorized usage, and how to create psychological safety so people don’t hide their AI use.
Then there’s the overuse problem. Some people let AI completely remove their personality from their work. Everything becomes the same bland middle. Same tone. Same structure. Same as every other organization. That’s a real danger on the other side. Managers need to know how to spot when someone’s using AI as a crutch versus a tool, how to give feedback that encourages smart use rather than avoidance, and what good versus problematic AI use looks like in your specific context.
The adoption gap is where things get really difficult. You’ll have team members who are AI-native, who’ve figured out how to save 10-20 hours a week, whose throughput increases 40-50%. Then you’ll have team members who are skeptical, scared, or struggling, who barely use it.
This creates massive challenges around performance evaluation when some people have AI leverage and others don’t. Do you push more work onto the AI-native employees? How do you close the gap without forcing the skeptical ones? How do you ensure those not adopting don’t get left behind? Managers need frameworks for setting expectations across different adoption levels, how to have developmental conversations with low adopters, how to ensure AI-native employees don’t burn out from extra load, and principles for fair performance evaluation in this transition.
Time What Is Time?
If people find 10% efficiency in their day with AI, what do we do with that time?
This requires a philosophical discussion that most organizations avoid. Does this mean we hire fewer people but have each person do more? Do we use this to actually reduce workload and address burnout? How does this connect to our mental health and burnout statistics?
Managers need guidance. They can’t make these calls in isolation. But they’re the ones who have to implement whatever decision gets made. You need to cover in training what the organization’s actual intention is - and be honest about it - how to communicate this clearly to teams, how to spot when efficiency gains are just creating more stress, and when to push back on “just do more with the time.”
The Employee Experience Problem
If I’m on a team where my manager loves AI, I get one experience. If I’m on a team where my manager is skeptical, I get a completely different experience. We’ll never get 100% consistency. That’s impossible. But we need similar baseline conditions.
This requires clear organizational expectations for how managers should lead with AI, consistent messaging about what good adoption looks like, regular calibration between managers on how they’re handling this, and HR support when managers get stuck. You can’t just let every manager figure this out independently. The employee experience will fragment too much.
The Bottom Line
My firm belief is that you cannot skip training managers. You cannot just train employees and hope managers figure it out.
Managers need two distinct capabilities: using AI competently in their own work, and leading teams where AI adoption varies wildly.
Without this, your adoption efforts will fail regardless of how good your employee training is. Because managers are the layer that makes or breaks actual behavior change.
Coming in Part 4: How to actually design and deliver the training content, what works, what doesn’t, and how to make it stick beyond the initial enthusiasm.
If you’ve trained your managers using this approach? You’re positioned for Part 4 on scaling to the full organization.
If you haven’t trained managers yet? Do it. Everything else fails without it.
And need help putting this all together? This is what I do. Reach out.
This rings very true. I recall a client CEO saying, "Here is what Perplexity came up with" and that set the tone for the whole company becoming AI native. It gave all the managers the permission they needed, and then the rest of the company followed.