Classify
What LLMs can do, where hallucinations begin, and why context, data, and prompting change the quality of the outcome.
For C-suite, entrepreneurs, and owners
Understand AI, to make better decisions.
Strategic AI sparring for leaders who already use LLMs and want to go deeper — to understand where AI is reliable, where it misleads, and how it can support stronger steering processes and more robust decision foundations.
Understand LLMs·Sharpen decisions·Improve steering
Starting point
Many use LLMs. Few understand what they are steering.
ChatGPT, Claude, and other LLMs are already part of daily work in many companies. What is often missing is not access to the tool, but a robust understanding of how AI works, where it holds up, where it misleads, and how it should be embedded into decisions, processes, and responsibilities.
That is where I work: between business model, leadership, and technology.
Steer
Where AI truly belongs in leadership, prioritisation, and decision processes.
Focus
Which use cases carry real weight and which ones merely create activity without impact.
Who this is for
Not for beginners. For leaders with first-hand AI experience.
You already work with ChatGPT, Claude, or other LLMs and want to move from tool play to judgement, clarity, and steering capability.
C-suite
Wants to place AI in a way that leads to better leadership, risk, and investment decisions.
Entrepreneurs
Want to understand where AI creates leverage and where it only produces hype.
Owners
Want stronger foundations for strategic choices, responsibility, and prioritisation.
Offer
Five formats. One goal: more clarity, better decisions.
The work
From first conversation to decision groundwork — AI sparring in four steps.
Stance
In the end, common sense still wins.
I work with AI, but I do not sell magic tricks. AI is only useful when its mechanics, limits, and risks are understood. That is why my approach is transparent, documented, and tied to concrete decision problems — not to tool hype.
AI can prepare, compare, and accelerate. Responsibility, prioritisation, and judgement stay with people.
To the Denkwerkstatt →White box over black box
You see the outcome and the path that led there.
Common sense over hype
Not everything that can be done with AI makes economic sense.
Responsibility stays human
AI does the groundwork. Decisions, prioritisation, and accountability remain with you.
Packages
Each package stands on its own — ONE concrete goal
Which one? We decide together. Individual. Tailored case by case.
Frequent questions
Asked most often.
About sparring
A training delivers content to a group. Sparring is individual and starts from your concrete use case — typically 1:1 or with a small leadership circle.
No. If you know your business model, that is enough. I translate the technology.
The analysis package starts at EUR 3,800. Anything beyond is individually priced, depending on scope and cadence.
Yes. First calls are usually over video. From format 02 I mix remote and on-site.
AI champions are employees who show a particular interest in new technologies such as artificial intelligence. They usually already have some experience — sometimes more — and can therefore more easily grasp how AI works. These people are highly sought-after and well-paid on the job market. → Use this advantage you already have in-house.
That is the beauty of AI. You do not need to be a technician, programmer, computer scientist or developer. ANYONE who takes an interest in this technology has the option to learn it. The mere fact that you are reading these lines shows that you are among the front-runners in your industry. Use that head-start.
What do your employees complain about most often? There you go — you now have your first lead.
Privacy & compliance
The standard version of ChatGPT is not suitable for company data. ChatGPT Enterprise, Microsoft Copilot with your own tenant, or local models are. → What you need, we clarify in sparring.
Depends on the AI and on the data processing agreement (DPA). Rule of thumb: no personal data without a DPA. No confidential content in free consumer tools.
Depending on the provider, somewhere between "nowhere" and "in the training dataset". Read the enterprise terms — or ask me.
Yes. Without clear ground rules, your employees reach for consumer tools. That is the biggest data-protection risk — not AI itself.
Cost & value
It depends on your biggest pain point. Some save 40% processing time, others build new services. Without analysis, every number is marketing.
How much time do your employees spend on routine tasks that cost time but generate no return?
Reducing opportunity costs is, from my experience, the biggest lever. You free up your employees so they can focus 100% on the job you originally hired them for.
Two further levers: quality jumps and new revenue streams. In sparring we calculate the biggest lever for you — before you invest in tools.
First: time. Anyone unwilling to invest time — their own or their employees' — is not ready for the coming AI transformation and will lose market share.
What matters is what it yields. If the value is less than the medium- to long-term increase in revenue, margin and profit — I'll tell you first. I only take on mandates that are meaningful and create real value.
As an anchor: the analysis package starts at EUR 3,800. Anything beyond is individually priced.
Especially. Small teams benefit most because every employee is a potential lever. Large corporations have fundamentally different requirements and strategies. Decisions take longer — SMEs often benefit from direct decision-making processes, and therefore from greater speed.
Rollout & implementation
A first productive use case: 4–8 weeks. Real transformation: 6–18 months. Anyone promising "full digitalisation in two weeks" just wants to sell you tools.
Real transformation happens from within. In the team. Those who build the future together carry these decisions substantially.
Both at the same time. Strategy without pilot is paper. Pilot without strategy is an island. Sparring connects them.
No. AI today gets along fine with Excel, Outlook and PDF. Modernisation is a side-effect — not a blocker.
The one that solves your pain point. ChatGPT, Claude, Copilot, Mistral or Lovable — all have strengths. Tool and vendor selection comes AFTER the use case, not before.
For most SMEs, cloud is the better choice — faster, more up-to-date, cheaper. But use cases and usage frequency make the difference. Every model has different strengths, and you don't always need the newest or fastest. In some cases a small local model running on your own server is enough; in others you need a powerful data centre.
Hallucination & trust
No. No model is — at least so far — error-free. Only within the right context and with the right controls. AI can be convincingly wrong. Those who know this use it safely. Those who don't have a problem.
The AI invents content that sounds plausible — but is wrong. Sources, numbers, quotes. Recognisable only if you know the topic. That is why domain expertise remains essential. Anyone who does not know what they are talking about cannot check what they are reading.
Three indicators: overly smooth wording, unverifiable sources, overconfidence on niche specialist questions. → A counter-prompt with "give me three sources" helps — unless you skip checking whether those sources actually exist. → See "What is an AI hallucination?".
The responsibility remains with you. AI is a tool, not a decision-maker. Those who ignore this still bear liability — then without an excuse.
People & team
Fear comes from uncertainty. Clarity comes from trying it out. First hour with AI in the team → fear drops by 70%. Experience, not statistics.
Some. Many will change. Those who understand AI become more sought-after. Those who ignore it become more replaceable. Simple, but uncomfortable.
The goal of any entrepreneur should not be to replace employees with AI. The goal should be to take tedious, mind-numbing routine work off their plate so they can once again focus on the job they were actually hired for. In most cases this automatically lifts revenue, margin and profit over the medium to long term.
Not with a two-day seminar. With real use cases from your own daily work, where the team learns — learning by doing. That is exactly what sparring is. In the process you identify your AI champions. They are the ones who will carry and drive the unavoidable transformation for you.
No. Two or three AI champions per department are enough. They pull the rest along.
Use cases & strategy
Summarising research, translating texts, drafting emails, analysing tables, finding patterns in data. Strategic scenario planning and much more. BUT don't let anyone sell you a story!
Not every task has to be replaced by AI. Often a classic database-query automation is enough. Especially tasks based on deterministic numbers and queries — you don't need AI for those, there are other, sometimes cheaper solutions.
Final decisions. Personal conversations. Anything where liability, ethics or empathy are central. AI delivers the pre-work. The decision stays with the human. A good example: these FAQs and this website were created with AI's help. Even so, I read every page, every line personally and adapt, polish, reword the content. Does that make me immune to mistakes? No — because I too am only human.
An agent is an AI that executes multiple steps on its own — not just answers. Very useful for recurring workflows. Overkill for one-off questions.
Parts of it, yes. Entirely, no. What pays off is the cyclical, recurring portion: preparation, classification, drafting. Sign-off stays with the human.
Regulation
European law, gradually entering into force since 2024. Depending on the use case (from low to high risk), different obligations apply. Most SMEs fall into "low" — but documentation is needed by EVERYONE.
Yes. Not just because of the EU AI Act — also for auditability and insurability. Those who don't document lose when disputes arise.
First conversation
Clarity first. Then decide whether it makes sense.
In 60 to 90 minutes we clarify what you are currently working on, how deep you want to go, and whether a meaningful mandate should follow from it.