I’m Nastassia

A lot of my work sits where a wrong answer costs more than a bad review: cell and gene therapy supply chains, legal and compliance AI, pricing data sold to carmakers. So I design for traceability, clear permissions, and answers that show where they came from.

I started designing AI products in early 2023, a few months after ChatGPT turned AI into something companies could buy. Policy AI for compliance teams came first. By July 2024 we had AI Agents live with customers, each one specialised in a single regulation, before “agents” became the word everyone uses. Then a Contract Agent for legal teams.

Designing AI agents since before they were called that.

AI changed how fast we can build, not what’s worth building. I focus on what decides whether people trust it at work: answers grounded in their own documents, sources you can check, feedback on every answer, and clear rules on who sees what.

I go to the source. I sit in the prospect calls myself instead of reading the notes afterwards, write down the team’s assumptions before research, and test with the cheapest thing that answers the question: a smoke test, a mocked page, a prototype used as a sales demo. Sometimes the best result is killing an idea early.

I’ve owned roadmaps at Babbel and Körber Digital, and stepped into the PM seat when a team needed someone to steer. I prioritise from evidence (plus a healthy dose of judgement), and I bring engineers into discovery early, so they own the outcome and not just the tickets.

Products with many parties and many rules: 4 organisations sharing one patient record, forecasting and pricing teams who must not block each other, AI answers that depend on who’s allowed to see what. My job is to make each new piece fit the whole, so the product still feels like one product.