First case won using Andri

First case won using Andri

Last week, Ronald Zwiers won a case. Nothing unusual there—he's been practising law for years. What was different this time: Andri helped prepare it.

This wasn't a demo or a proof of concept. It was a real dispute, real opposing counsel, real stakes for the client. And we learned more from this one case than from months of building in isolation.

What actually happened

The case involved a contract dispute with a moderately complex fact pattern—several years of correspondence, multiple parties, and a legal question that wasn't straightforward. The kind of matter where you'd normally spend a full day just getting oriented in the case file.

Ronald used Andri to:

  • Map the fact pattern across thousands of pages of documents, pulling the key communications and dates into a timeline
  • Find the relevant jurisprudence on the specific legal question—not general case law, but decisions with similar fact patterns from Dutch courts
  • Draft initial arguments that he then refined based on his experience and judgement

The time saving was substantial. What would typically take most of a week condensed into two days of focused work. But more importantly, the quality was there. The arguments were well-grounded, the citations were accurate, and the strategy held up in court.

What we learned

Building legal AI in a vacuum teaches you one thing: what sounds good in theory. Building it alongside actual legal work teaches you something else entirely.

Lawyers don't want "AI-generated" arguments. They want a first draft that's good enough to refine. The difference matters. An AI that tries to replace legal judgement gets ignored. An AI that handles the tedious parts so the lawyer can focus on strategy gets used.

Citation accuracy is non-negotiable. We already knew this, but seeing it in an actual case filing reinforced it. One wrong citation, one outdated precedent, and credibility is gone. Our multi-stage verification exists for exactly this reason.

Context is everything. The same legal question has different answers depending on the specific facts, the jurisdiction, the court, even the judge. Andri's value isn't just finding law—it's connecting that law to the specific matter at hand.

What's next

This was the first case. Since then, we've been refining based on what Ronald and other lawyers in our pilot programme have shown us. The system gets better with every real-world use.

If you're curious about what Andri can do for your practice, try it. We're still early, but the foundation is solid—and it's been tested where it matters.


Read also: what we learned from our pilot with Blokziel, how LawBeam turned 100 hours into 10 minutes, and why agentic reasoning is the only path to production legal AI.