AI in client mandates
How I work with artificial intelligence.
In my work, artificial intelligence is a tool – not the mind behind it. This page sets out what I use it for, what I do not use it for and what happens to your documents. So that you do not have to ask.
The principle in one sentence
Analysis, decisions and responsibility rest with me. AI speeds up work steps that would have to be done anyway – it replaces neither judgement nor experience.
- Research and pre-structuring of material that I then check myself
- Wording variants and shortened versions for reports, presentations and website texts
- Reviewing large volumes of text, such as minutes or series of key figures
- Creating and editing individual image and graphic elements for this website
- Professional assessments, recommendations and decisions in the mandate
- Evaluations of people, performance or job applications
- Key figures, project descriptions and statements about results
- Anything that goes out without my review
What happens to your documents
This is the question that comes up first in banks. That is why it comes first here.
- I do not enter confidential documents from mandates into open, freely accessible AI services.
- Where AI tools are to be used in a mandate, I set out the purpose, system and scope in writing beforehand and obtain your approval.
- If your institution has its own rules on the use of AI, yours apply. No discussion.
- I do not process personal data of your employees with AI tools.
- I disclose which systems I use on request and record this in writing in the mandate.
In short. No document from your institution goes into an AI system without your consent. If in doubt, I ask rather than act.
Responsibility stays with me
A tool relieves me of work, not of liability. Every text, every figure and every recommendation that bears my name has been checked by me. If something is wrong, it was me – not the machine.
This also applies to this website: professional statements, project descriptions and key figures come exclusively from my own experience and verified sources.
How I label AI use
Since 2 August 2026, the transparency obligations under Article 50 of the EU AI Act (Regulation (EU) 2024/1689) have applied. This is how I handle them:
- Image, graphic and video elements on this website that were created or edited with generative systems are labelled via the general notice in section 11 of the legal information (in German).
- Articles carry a note at the end stating that AI-supported tools were used in their preparation and that the author is responsible for checking the content.
- Where AI generation is not obvious and the general notice is not sufficient, I point it out on the spot.
- Disclosure beats waiting. If in doubt, I would rather label once too often.
Why I publish this
Because the question comes up anyway. Compliance departments now check how commissioned third parties handle AI. Anyone without an answer loses time in the procurement process – or the mandate.
I handle it here as I do in a project: straight talk beforehand instead of explanations afterwards. Do you have requirements that go beyond what is described here? Tell me in the first conversation and we will put them on record.
As of: 19 August 2026. This page supplements the legal information (in German) and the privacy policy (in German).
And how do you handle it?
This page answers questions you would ask me. The same questions are worth asking internally – and of your service providers, intermediaries and cooperation partners:
- Do you know which AI tools are actually in use in your institution – including those that nobody has approved?
- Is there a written rule on which data may go into which system? And does everyone who needs it know it?
- Who checks the result before it goes to the customer – is this role named or tacitly assumed?
- How do your partners handle it? Have you asked them, or are you assuming?
- What do your contracts with third parties say about it?
- If your supervisory authority asks tomorrow: can you prove who made which decision?
If you hesitate on more than two questions, that is not a criticism – it is currently the norm in many institutions. But it is a gap. And gaps are better closed beforehand than in the audit report.
Where AI reaches its limits in transformation:
Your next step
Questions about AI use in your mandate?
Bring them to the first conversation. Thirty minutes, free of charge – and if your institution has its own requirements, we will put them in writing straight away.