AI solutions

By calculating it before the project rather than after. We start every AI initiative with a feasibility check: how often the task occurs, how long it takes by hand today, what the error rate is, and how good the data is. From that follows whether the effort pays back within a reasonable period. If simple automation without AI solves the same task more cheaply, we say so and advise against the AI project.

We assume it will and build accordingly. Every result gets a confidence score, and anything below the threshold goes to a person instead of onwards automatically. For document processing we always show the source passage in the original, so a check takes seconds rather than minutes. All decisions are logged, so it stays traceable afterwards why a result came out as it did. We do not build a black box that issues results beyond question.

For most operational applications not much, but not nothing. The regulation classifies systems by risk. Document processing, text preparation and internal assistance usually fall into the lowest tier with transparency obligations: people must be able to tell they are dealing with an AI system. It gets stricter for recruitment, credit scoring and similar cases. We classify your initiative before the project and tell you what follows from it.

Usually four to eight weeks from order to use. It starts with the feasibility check, which shows whether the project pays off and whether the data is good enough. The duration after that depends mainly on how many document or enquiry types the solution should handle and which systems are connected. We start with one clearly defined use case and only extend once it works day to day. You receive a binding schedule with the fixed-price quote.

Small adjustments during the project are included in the fixed price. We work in short steps and show you the current state continuously, so changes surface early rather than at the end. If a new requirement comes up that goes beyond the agreed scope, you receive a quote for it first, also at a fixed price. You decide whether it is built now or in a later stage. Nothing gets more expensive without your prior agreement.

Besides running it, the AI providers charge a fee per request, depending on how much text is processed. These fees run on your own company account with the provider, which we set up together with you, and are billed to you directly. How high they are depends mainly on volume. For clearly defined decisions, such as sorting or routing, there are now specialised models that cost a fraction. We give you the running costs after the feasibility check, before launch, together with the calculation of whether it pays off.