Teaching People to Use AI Is the Easy Part

← Back to Writing

The gap most programs don't see

Most organizations' AI programs stop at tool adoption: teaching people to prompt faster, draft faster, summarize faster. That work is necessary. It also isn't what decides whether the transformation sticks.

The real constraint shows up a level higher, in whether leaders know how to run a team whose work is changing shape underneath them. That's a leadership-capability gap rather than a training-content gap, and most curricula never get built for it. They get built for individual contributors, with a "leadership module" bolted on near the end almost as an afterthought.

My approach doesn't start with curriculum. It starts the way I've approached every transformation I've led: understand how the work actually gets done, find where the friction and risk live, build the governance and reinforcement system required for real adoption, and give leaders the judgment to manage a team whose capacity is shifting in real time. Learning is one component of that system, not the whole of it.

A leader-first progression, not a course catalog

Four stages, each built around what a leader needs to know and do at that point in their team's AI maturity, rather than a generic literacy-to-expert ladder aimed at everyone equally.

• Foundation. Leaders build real fluency with the tools their own teams use, so they can coach from experience instead of reciting policy.
• Workflow judgment. Leaders learn a repeatable way to look at their own team's work and separate what should stay human, what AI can assist, and what's a real candidate for automation, applied against their actual workflows rather than a hypothetical example.
• Capacity leadership. The stage most programs skip entirely. More below.
• Governed autonomy. Leaders learn what responsible delegation to AI systems looks like in practice: clear ownership, clear boundaries, human review where it matters. Confident leaders, not unmanaged sprawl.

The stage that gets skipped

When AI removes low-value work, the question that matters is what the freed-up capacity actually becomes. Left unmanaged, it turns into one of three things, and each needs a different leadership response.

• Waste. AI output that looks productive but adds review burden instead of removing it. The fix is naming it and setting team norms, the way early email needed etiquette before it was usable.
• Necessary overhead. The real, upfront cost of making a team AI-capable. This has to be planned and resourced, and treating it as a failure when it shows up is a mistake.
• Real breakthroughs. The unplanned discoveries that come out of a team actually using these tools well. Leaders have to be watching for this, or it gets missed because it never shows up on a normal capacity report.

Catching which of these three is happening on your team in real time is an observation skill, not a tool skill. It's almost never taught, and it's what separates a leader who's managing AI adoption from one who's hoping it works out.

The resistance nobody names

AI resistance shows up in senior employees worried about being displaced. Just as often it shows up in the newest hires, people early in their career who are genuinely unsure whether AI-assisted work will be respected, whether leaning on it will stunt the skill-building their career depends on, or whether the effort behind their work still counts for anything once a model can produce a draft in seconds. Both groups are protecting something real, and it isn't the same thing.

Framing this as "is AI-assisted work fake" undermines the adoption you're trying to drive. The better question is what human signal we have to consciously protect as production gets cheaper: judgment, accountability, care, relationship. Leaders who can answer that clearly, and say it out loud to their own teams, get durable adoption. Leaders who don't get surface-level compliance and quiet resentment.

Why this has to come from the top

A culture of innovation gets built by what leadership visibly does, not by training content. If executives approve the initiative but don't personally use the tools, employees notice the gap immediately, and the program reads as "this is for you, not for us." Sponsorship has to mean leaders modeling the exact behavior the program is trying to teach, not just funding it. Make that a named success criterion rather than an assumption.

Grounded, not theoretical

• Built from real transformation work. I led the design of an AI adoption program as the first-ever CIO of a 100-plus-year-old organization, with governance, champions, training, and executive sponsorship built together instead of training as a standalone workstream.
• As a Microsoft Certified Trainer, I actively teach the certification tracks this framework draws on: Microsoft's AI transformation curriculum for directors, VPs, and executives, and the technical courses behind agent design and agent evaluation. The governance and measurement layers aren't hypothetical.
• Grounded in current research. Gartner's own data shows AI use creates real work friction alongside the savings, and that unmanaged AI intensification is now a measurable driver of burnout in an organization's best adopters. Both are design principles here, not edge cases.

The one ground rule

This is a framework, not a prescription. Every organization's culture, drivers, and strategic priorities are different, and a learning-and-change program that isn't adapted to those specifics doesn't survive contact with a real organization. Before I build anything from this, I sit down with the actual stakeholders to understand the business drivers behind the request: what's forcing the timeline, what's already been tried, where resistance really lives. The shape here is reusable. The content is meant to be built around what your organization actually needs.