Gavn Software

We started Gavn because we had both watched the same thing go wrong.

Andreas was building secure infrastructure and production machine learning, most recently in a security operations centre where he automated 55% of incoming phishing cases. Niels spent seven years on flight-planning infrastructure, where a system failing at the wrong moment is not a bug report. Different halves of the same problem: AI that has to run inside systems people depend on.

What we kept seeing was distance. The person who understood the problem sat three handoffs from the person writing the code, and what shipped was a reasonable interpretation of a reasonable interpretation.

So we built a company with no distance in it. Gavn is two engineers. You talk to the people writing your code, from the first conversation to the day the system runs in production.

We are engineers before we are an AI company, and the order matters. Most AI projects fail on the foundation rather than the model. We design from the failure cases first, strip out every part that is not earning its place, and make AI prove it beats a simpler solution before we build it. Sometimes it does not, and we say so. That is a cheaper conversation than a pilot that demos well and never ships.

We work three ways. A two-week AI Readiness Assessment that examines your systems and workflows and returns a prioritised, costed plan, yours with or without us. Workshops of one to three days on your own codebase and backlog, not toy examples. And the engineering itself: agents built into production software with retrieval, guardrails and evaluation harnesses; custom machine learning for forecasting, classification and anomaly detection; and the infrastructure underneath in Python, Rust, .NET and Azure. GDPR, EU data residency and AI Act obligations are in the architecture from day one, not a review at the end.

The clearest example is PDC StaffPlan, a Danish workforce planning product. We built an AI agent inside it, grounded in 20,000 pages of manuals. Four months from specification to production. It runs across seven of their products for more than 10,000 users. Their Head of Development put it this way: "These guys are sharp. The agentic approach has reshaped how StaffPlan competes in its market."

Both of us hold an MSc in Computer Science and Engineering from DTU and still teach there. Our name comes from the university motto, til gavn for samfundet: for the benefit of society. We took it because we meant it.