Do the maths first, then automate.
The question is rarely whether AI can do something. The question is whether it pays off for your business, who runs it afterwards and what happens when the result is wrong. That is where the assessment begins.
Why I can judge this
No software is sold here and no provider is referred. I earn nothing from the introduction of a particular tool. On top of that comes a combination that is rare in this field: judging the technology and the numbers behind it.
- Diplom-Ingenieur (FH) in microelectronics and information technology, Executive MBA from the Technical University of Munich
- University certificate „Generative KI produktiv nutzen“ (using generative AI productively) from Hochschule Hannover, including the legal module on the AI Act, GDPR and copyright
- Daily productive use in my own consulting practice, not as a demonstration but as a tool
- More than twenty years in industry, from aircraft maintenance to managing a hardware manufacturer with its own production
Whoever sells you a tool will rarely tell you that you do not need it.
How an assessment works
Collect use cases. Not what would be possible, but where your business actually loses time or where errors occur.
Check the technology. What can be done reliably with today's technology, what only under supervision, and what is marketing.
Calculate the economics. Effort for the introduction, running costs, the data required, and who checks the result in the end. Only then is it clear whether something pays off.
Name the legal guardrails. Where data protection, the AI Act or copyright limit the implementation. This is an assessment, not legal advice.
Recommendation with an order. What you start with, what can wait, and what you had better leave alone.
If none of it pays off, you will hear that. A result that says „in your case only one of these five use cases pays off at the moment“ is a good result. It has saved you the other four.