AI Competence Comes From Doing, Not From Knowing

You can book your team the best AI training programme tomorrow. Six weeks later, nobody will be using the tools. Not because the training was bad. Because there's a gap between "I understand what AI can do" and "I use AI every day for my work" that no webinar can close.

Building AI competence in your team is harder than it sounds. 70-80% of all AI initiatives fail, and the most common reason isn't bad technology. It's missing change management.

What goes wrong in most companies

The pattern is always the same. The company buys licences, IT rolls them out, HR organises a training session. Everyone says it's "really exciting". Two weeks later, nothing has changed in the day-to-day work. The access is there. But nobody knows exactly what to use it for.

This isn't about a lack of willingness. It's because the setup is structurally wrong. Knowing about AI and applying AI to your own, concrete problems are two different things.

In 2025, 42% of companies scrapped most of their AI projects. The budget was there. The tools were there. What was missing: involvement from the people who were supposed to actually use them.

The difference between training and competence

Training teaches what AI can theoretically do. That has its place. But AI competence only develops when someone takes a real problem from their daily work, works on it, and sees at the end: it worked. I can do this again.

Training builds knowledge. Competence comes from applying it to your own problems.

The difference in practice: a team that's been through training knows AI can summarise text, analyse data and draft documents. A team that's built an AI prototype for its own sales problem knows something more: how to replicate that the next time a problem comes up.

Why AI training in companies often falls short

Training fails in two places.

First, it's generic. "50 things you can do with AI" isn't competence development. It's a list of options with no context. The sales rep is left wondering: so what do I actually do with this in my job?

Second, training misses the crucial final step. Every good learning experience ends with a transfer step: I learned something and applied it to something of my own. Webinars end with a Q&A and a PDF to download. That's not transfer.

According to recent surveys, only 13% of employees have actually received AI training, even though 77% of employers are planning reskilling. The topic is on the agenda. Execution is lagging behind.

What works instead: three principles

1. Real problems, not example tasks

Let the team bring their own problems. The process that eats three hours every Monday. The report nobody reads but everyone has to produce. The proposal phase that repeats itself every single time. When that's the raw material, learning stops being abstract.

2. Build it yourself instead of watching

A team that's built an AI prototype for its own problem understands more than a team that's watched ten demos. That's not an opinion. It's the explanation for why hands-on involvement leads to higher adoption than training does.

3. Leadership sits at the table, not in the stands

In 2026, 75% of CEOs describe themselves as their company's primary AI strategist. At the same time, many delegate the operational work to a Head of AI and watch from the sidelines. That sends a clear signal to the organisation: this isn't a priority.

When a leader sits in the workshop and works on a problem themselves, the dynamic in the room changes. The team sees: this matters enough that my boss is putting their own time into it.

AI adoption in the German Mittelstand: the ownership principle

People defend what they've built themselves. What gets delivered from outside ends up in a drawer.

That's the core of the ownership principle. When a team builds its own AI prototype, it has a real stake in making sure that prototype actually gets used. Because it's their project. Because the problem was theirs. Because the solution came from the team, not an outside agency.

This isn't a workshop gimmick. It's the difference between AI adoption and AI investment with no return.

The AI Design Sprint builds this principle into the process. Teams identify their own pain points. Sales works on sales problems, production works on production problems. At the end, there's a working AI prototype the team built itself. The knowledge stays in the company. So does the motivation.

I show what a broader AI strategy built around this approach can look like in my article on AI adoption as a change management task.

What you can do differently for your next AI initiative

Before you approve the next training budget, ask one question: what should concretely be different afterwards?

If the answer isn't specific, the training isn't an investment. It's an expense.

Specific looks like this: the sales team applies AI to at least one of their recurring processes over the next four weeks. And afterwards, they have something they can show.

Your next step

I help teams build AI competence through real application, not slide decks. If you want to know what that could look like in your company, book a free Discovery Call.

S&P Global 2025: 42% of companies scrapped AI projects
Conference Board 2026: 75% of CEOs act as primary AI strategist
McKinsey/DataCamp surveys: 13% of employees received AI training, 77% of employers planning reskilling

Back to overview