Founder-run. Pre-award AI for the people who carry the largest decisions in industrial sourcing.
At most industrial companies, the largest line item by far is what they buy from suppliers. Three to five times the R&D budget at a car company. Six times the headcount cost.
And the decision of which supplier — the actual moment of evaluation — happens in a spreadsheet.
In the last four years we've watched awards north of a hundred million euros where decision prep meant emailing the bid PDFs to India to be pasted into Excel, the spreadsheet shipped back to Germany, then a team of engineers assessing each line by hand — against yet another stack of PDFs. We've been asked to rebuild decisions customers had already made, because nothing was written down. We've seen customer-supplier relationships break over phrases like "the industrial PC must be from a famous brand" — six months in, no one can agree what that sentence ever meant.
Today, most of an expert's time in this process goes to that kind of preparation — moving documents, transcribing requirements, comparing lines against lines. The decision itself, the part the expert is paid for, gets what's left.
The people inside the work know it. They can't fix it from where they sit.
And whatever fixes this has to live where the data lives. Industrial customers don't ship bid books, supplier pricing, or award rationale to a vendor's cloud to be processed. The model has to come to the data — inside the customer's perimeter. Not the other way around.
Perspix is the response.
The decision is the work. Everything else — the prep, the data shuffling, the reconstruction six months later — is around the work. We're here to flip that. Most of the expert's time on the decision itself. The rest, review.
Sourcing without friction.
Six commitments that shape how we work — and how we serve customers.
We co-develop strategy with each customer. Solutions built for the specific shape of your problem, not the average.
Dynamic structures. Short decision paths. Projects move quickly and we don't become the bottleneck.
Respectful, prudent handling of data, people, and the environment. Security and ethics are foundational, not bolted on.
Deep know-how in AI and in requirements documents. The best possible solution, made easier to choose.
Agile response to change. Processes adapt to market and customer needs continuously.
Openness and clarity in all processes. Clear communication, traceable procedures, decisions you can defend.
Geschäftsführer of Perspix.ai GmbH (HRB 750649). Runs the operator side of Perspix and customer engagement. Demos on this site are run by him or by a Perspix engineer — not by an SDR.
Co-author on all four of Perspix's published machine-learning papers, at NeurIPS (2024, 2025), ICLR (2024 workshop), and OPT 2025. See Research for the publication list.