What the question usually looks like
Somebody in sales is already pasting enquiries into a chat tool. The managing director has been asked at the trade association what the company is doing about AI. The honest answer is: nothing has been decided yet. The Bitkom Research survey “KI in der deutschen Wirtschaft 2026” puts figures on that state: 30% of the companies without AI name missing use cases as a barrier, and costs that cannot be estimated are named by 54% of users and 63% of non-users.
That is a selection problem, not a technology problem. The AI projects that fail later rarely fail on the model. They fail on the process that was picked: too rare to pay back, too particular to describe, or built on data that does not exist in the form the plan assumed. That choice is made before anything is built, and it is the cheapest place to get it right.
How the assessment day runs
One working day, at your site or online. In the room: the people who do the work every day, one person who is allowed to decide, and us. Not the IT department on its own, and not management on its own.
In the morning we walk through real cases rather than the process diagram. A quote from last week, from the enquiry in the mailbox to the PDF that went out. An incoming invoice, from the scan to the posting. A service request, from the call to the closed ticket. What we are after are the exceptions no diagram carries, because they decide whether a build holds up.
In the afternoon every candidate gets four figures, and they come from you: how often it occurs, how long it takes, how often it goes wrong, and what it costs when it does. Alongside that we record which data sits where, which systems hold it, and which AI tools are already in use in your business, approved or not. We do not estimate a business we have known for six hours; we write down what you know and what nobody knows yet.
Then the write-up at our end.
What you hold at the end
- A prioritised roadmap. What to tackle first, what can wait, and what to leave alone, with the effort and the expected return per step written as a range with its assumptions, not as one number.
- The exclusion list. The processes where the effort does not pay off, and why not. It is the part that saves the most money.
- The data question per candidate. What exists, what is missing, what would have to be maintained first.
- The EU AI Act risk class per candidate, with purpose, data sources and limits documented, so the compliance work starts before the build rather than after it.
- A written basis for the first build. Scope, systems involved, dependencies and cost, detailed enough that another supplier could build from it.
A worked example
Take a machine builder in the Düsseldorf area with a small sales team. Three candidates come up on the day: quoting, incoming invoices, and technical enquiries from the website.
Quoting turns out to be the strongest. It is frequent, it ties up the most experienced people, and every quote draws on the same three sources: the ERP price list, the last comparable quote and the enquiry itself. It goes first, as an agent that drafts and a person who sends.
Incoming invoices look attractive, but the ERP already has a capture module nobody switched on. That goes on the roadmap as configuration, not as an AI project.
The technical enquiries are dropped. Too few per month, and the answers live in the head of one engineer rather than in documents. That is a documentation problem first, and it goes on the exclusion list with the reason written next to it.
What you need to bring, and when the day is not worth it
You need the people, not the documents. A process manual is useful; the person who ignores it on a busy Friday is essential. You need one decision maker in the room for the whole day, and you need to be prepared to hear that a favourite idea is not worth building.
The day is not worth it if you already know your first use case, can describe it, and the data for it exists. Then say so in the introductory call and we go straight to the build. It is also not worth it for a business with one process and no systems to speak of; there the answer fits into the call.
How it differs from the other services
The assessment decides. Every other service builds, connects or trains, and each of them assumes that the decision has been made. AI agents and process automation take on the work itself. Knowledge systems answer questions from your documents. Integration is the connection into the systems you run, data and analytics puts the numbers behind it in order, enablement equips your people, and custom software fills the gap where no product exists. Which of those you need, in which order, is what the roadmap says.
The next step
The free introductory call takes thirty minutes. In it we find out whether an assessment day makes sense for you at all, or whether the answer is shorter than that.