01 THE CHALLENGE / WHY
The Tool Was Ready. The Organisation Wasn’t Quite Yet.
One morning, an email landed in everyone’s inbox: “You can enter Claude via SSO.” IT had done something bold — and right: give everyone access from day one. No gatekeeping, no long pilot phases. Within hours, every appliedAI employee could open Claude Cowork — a tool that does something fundamentally different from the AI tools the team had been using before.
Claude Cowork was not introduced instead of existing tools. appliedAI already uses meinGPT — a GDPR-compliant, Germany-based AI chat platform for secure enterprise use — for chat-based AI tasks, and runs local AI capabilities with NVIDIA infrastructure — and continues to do so for good reasons. Claude Cowork added a new category to that portfolio: agentic capabilities. Unlike a chat-based AI that responds to questions, a tool with agentic capabilities can read and edit files, execute multi-step workflows, and act autonomously on your behalf. That distinction matters — and it was not yet obvious to everyone.
Some people dove in immediately. Others weren’t sure where to start. Information was scattered and guidance was still catching up with the rollout. The enthusiasm was real — but so was the uncertainty. And beneath the surface lay a more structural challenge: many of appliedAI’s processes ran on implicit knowledge that nobody had ever written down. That knowledge gap doesn’t become visible until you try to hand a task to an agent — and realise you can’t explain it clearly enough for the tool to execute it reliably.
64% said AI success depends more on people’s adoption than on the technology itself. appliedAI was experiencing exactly that gap — and decided to close it deliberately.
IBM surveyed CEOs globally
02 THE TURNING POINT / WHY THIS CALLED FOR MORE THAN A GUIDE
A Forcing Function, Not a Handbook
Before the Führerschein was built, a governance framework was already in place: principles for how appliedAI approaches AI tooling, what is automated and what is not, and how the organisation thinks about responsible AI use. The Führerschein didn’t create that governance — it made it tangible and transferable for every employee.
A PDF or slide deck wouldn’t have worked. You cannot build intuition for a tool with agentic capabilities by reading about it. And there is a more fundamental reason: the most important skill in working with an agentic tool is not knowing which buttons to press. It is knowing how to think before pressing them. To delegate reliably to an agent, you first have to be able to describe a process clearly — its inputs, steps, and outputs. That turns out to be harder than expected.
No bot for bullshit. Sounds catchy. Turns out it’s harder than it sounds — because the moment you try to automate something, you realise how many processes run on implicit knowledge nobody ever wrote down. AI doesn’t let you skip that step. If you can’t explain a process clearly, Claude can’t execute it reliably. That’s not a flaw. It’s a forcing function.
appliedAI Change & Enablement team
03 THE SOLUTION / HOW IT WORKS
The Führerschein: 25 Minutes That Change How People Think
appliedAI’s Change & Enablement team built a mandatory 25-minute course that every employee completes before accessing Claude Cowork’s agentic features. They called it the Führerschein — the driver’s licence. The name is intentional: just as you wouldn’t hand someone the keys to a car without ensuring they understand the rules of the road, you don’t give someone access to an agentic tool without a shared foundation for using it well.
The Führerschein doesn’t teach how to use the tool. It teaches how to think before using it. The core lesson is one principle: understand and simplify a process first. Then automate. Never the other way around. Alongside that, the course introduces what makes a tool with agentic capabilities different from a chat-based AI — and where the real risks and opportunities lie.
Completing the Führerschein unlocks the full Cowork environment, including a set of pre-built workflows — called skills — that give Claude deep expertise in specific tasks. Skills are reusable instruction sets: install one and Claude knows exactly how to help, following appliedAI’s standards and processes. Over 10 skills come pre-installed, covering tasks like creating branded presentations or submitting use case ideas. A growing library of additional skills lives in appliedAI’s internal cloud, where employees can find, install, and contribute their own.
And while working in Claude Cowork, if an employee discovers something useful — a prompt that worked better than expected, a workflow that saved real time — they can share it directly with colleagues with a quick prompt. Claude formats the discovery as a structured post for appliedAI’s dedicated Slack channel #aai-role-model-community, where the whole team can learn from it.
Technologies & approach: Claude Cowork (Anthropic) · Large Language Model · Skills-based architecture · RAG knowledge base (internalAI) · Internal cloud skill library · Slack peer-learning community
What the Führerschein covers:
What makes a tool with agentic capabilities fundamentally different — and what that means in practice.
Understand and simplify first. Automate second. The Führerschein explains why “No bot for bullshit” is harder than it sounds.
Key safety rules for working with an agent — including how to read permission prompts, when to stop, and what to do when something doesn’t work.
04 RESULTS / WHAT CHANGED
From Scattered Start to Shared Structure
TABELLE
“The real shift happens gradually — through using the tools, making mistakes, and sharing what you learn.”
appliedAI, from the Führerschein course
04 WHAT THIS MEANS FOR YOU
The Technology Is Rarely the Hard Part
Organisations rolling out tools with agentic capabilities often encounter the same gap: access is broad, but understanding is uneven. Some people experiment enthusiastically; others wait for clarity that never quite arrives. Useful discoveries stay with individuals rather than reaching the team. And the processes that would benefit most from automation turn out to be the ones nobody has ever clearly defined.
The Führerschein approach is transferable: a short, mandatory, principles-first gateway that teaches process thinking before tool operation. Paired with a shared skill library and a peer-learning community, it turns individual capability into something the whole organisation can build on.
This may be relevant for your organisation if…
- You have rolled out an AI tool but adoption and quality of use remain uneven across the team.
- Your most important processes run on knowledge that lives in people’s heads — and you sense that AI will force you to write it down.
- Useful AI discoveries are not reaching the rest of your team, and the learning is not compounding.
06 LOOKING AHEAD
A Start, Not a Finish Line
The Führerschein is a foundation, not an endpoint. The skill library in the internal cloud grows with every employee contribution. The #aai-role-model-community channel evolves as people share what they find. And appliedAI’s approach to AI tooling continues to develop: make tools available, observe what people actually adopt, and decide strategically which solutions best address which challenges. Not a single vendor. A portfolio.
For topics as strategically important as knowledge management, appliedAI is building deliberately — with a conscious eye on avoiding vendor lock-in. How an organisation stores, retrieves, and reasons across its collective knowledge is too important to leave to a single provider. What that architecture looks like at appliedAI — and how it connects agents, data, and people across the tool portfolio — is the next chapter.
Stay tuned.