AI projects rarely fail on technology. They fail because the use case is vague, the data is not ready, nobody measures answer quality, or the legal team raises concerns just before launch. Our AI consulting therefore starts before building: which problem should be solved, how do we measure success, which data is needed, and which duties apply?
Assessing use cases
In an opportunity assessment we collect, with the business units, workflows in which a lot of text is read, written, sorted or searched: answering enquiries, checking documents, finding knowledge, drafting reports, transferring data from forms. Each case is rated for benefit, feasibility, data and risk. Some ideas turn out to be classic automation without AI, which is usually cheaper and more robust. We say so openly.
Chatbots and assistants on your own data
The most common productive use is an assistant that answers questions from your own documents: for customers on the website, for staff on the intranet or for support. Technically this is retrieval-augmented generation: the answer relies on retrieved passages and cites them. What matters is how documents are prepared, permissions so nobody receives answers from documents they may not see, and continuous measurement of answer quality.
Integration into existing systems
AI delivers most value where the work already happens: in the CRM, the ticketing system, the document store, Microsoft 365. We connect language models through APIs with clear permissions, logging and human approval for steps with external effect. Where agents should operate several tools, we rely on tightly bounded workflows rather than general autonomy.
Data protection, the AI Act and KI-MIG
For every project we clarify which personal data is processed, where the models run and which contracts with providers are needed. We choose model providers and regions so data can stay in Switzerland or the EU. Under the EU AI Act we examine the role as provider or deployer, the transparency duties of Article 50 and whether a high-risk case exists. In Germany, the Federal Network Agency has been the central market surveillance authority under KI-MIG since 29 July 2026, with BaFin responsible for the financial sector.
How we work
Innopulse uses AI daily in its own product portfolio, from content processing to assistant features in our SaaS products. We therefore know not only the possibilities but also the costs, the failure modes and the effort of reliable operation. A project ends with a documented solution, quality measurements and a team that can run it.