A workshop for the law's hardest documents.
ZAAN is a small, profitable, privately held company headquartered in Geneva and Shenzhen. We build software for lawyers — and for legal aid clinics, and for the people who never get a lawyer at all. The work is the same.
We started in 2022 with a question: why is contract review still done the way it was done in 1995? The honest answer was that nothing better had been built — yet. We've spent four years trying to answer that more carefully.
To make rigorous legal work available to everyone who needs it.
Half our team is paid by the firms who use ZAAN. The other half is paid to give it away — to legal aid clinics, public defenders and small-town solo practitioners across the world. We have no plans to change that ratio. We have plans to widen it.
Six things, closely held.
These shape the product, the hiring bar and how we spend our time. They were written in the first year and edited twice in four. They probably won't change again.
Precision before polish
A legal tool that's almost right is wrong. We ship slowly and check our work. Shortcuts the user can't see are still shortcuts.
Simplicity without loss
Every interface decision removes something. We remove weight, never depth. Restraint is a feature.
Useful to the underserved
We give our platform free to legal professionals in developing regions. It's not a marketing line — it's why half of us are here.
Write, then write again
Clear writing is clear thinking. A memo that reads cleanly almost always has a cleaner architecture underneath.
Quiet ambition
We don't announce what we haven't finished. The work speaks; we try not to.
A small, permanent team
We hire slowly for people who want to stay. Growth happens inside roles as often as between them.
Three people, one long argument.
ZAAN was started by a magic-circle lawyer, a research scientist and a product designer who'd built tools you've used. They argue often. They've never disagreed about why.
Amélie Roussel
Co-founder · CEOSpent eleven years at Slaughter and May before becoming convinced the work needed better tools — and started building them. Reads four contracts a week, on principle.
Kenji Tanaka
Co-founder · CTOResearch scientist at DeepMind, then Anthropic. Believes the model is the easy part — the hard part is making the reasoning legible to a senior partner who has 90 seconds.
Pia Lindqvist
Co-founder · Chief Design OfficerBuilt design systems at Linear and at Notion. Holds the line on restraint. Has been overheard saying "less, but precise" more times than anyone has counted.
42 people. And counting, slowly.
We hire roughly one person a month. Each one stays a long time. A representative slice of the people you'd be working with — engineers, lawyers, designers, researchers and the ops team who keep the doors open.
Four years, told quietly.
We don't tell origin stories well — we don't really like the genre. Here are the moments that mattered, with the marketing varnish scraped off.
A sublet office in Carouge, Geneva. The first month is spent reading contracts and writing eval sets. No code is shipped.
A boutique Genevan firm pays for a two-week pilot. The product is a Python script and a spreadsheet. The contract is for CHF 4,000. The team frames it.
CHF 6M from Stride.VC and Compound. Deliberately small. The team commits to never raising a round purely for headcount.
Three people, six months. Half the engineering team rotates through within the year. The cross-border M&A focus emerges from this office.
Mostly small firms in EU and SEA. Mostly from word of mouth. The team adds a single GTM hire — and only one.
The free tier for legal aid clinics ships. By year end it has more weekly users than the paid product. We consider this a feature.
CHF 28M led by Index. Used to build out research, deepen the in-house Counsel models and open a second engineering office. Headcount stays under 35.
The third surface — semantic search across firm history — goes live. Two customers cancel competing platforms within the week. We do not announce it.
Cash-flow positive on the recurring book. The team takes a week off, deliberately.
Cross-border M&A is now a quarter of revenue. Open Bar serves 4,200 paralegals in regions we cannot easily reach.
The first in-house model trained end-to-end for clause-level reasoning. It's small, fast, and good. The research team takes the rest of the month off.
Ten years of precedent, in three seconds. Multilingual citations and the trace view.
Three rooms, two timezones.
Most of the team works in Geneva or Shenzhen, with a small remote contingent across Europe and South-East Asia. The offices are quiet and well-lit. Phones rarely ring.
1206 Genève
前湾一路35号青年梦工场6号楼112
Quarterly off-sites in alternating offices.
Capital with long horizons.
We chose investors who don't ask us to behave like a rocket ship. They've all been founders themselves, they all understand the work, and they all read the contracts we send them carefully.
What others have written. Selectively.
We don't court press, and we won't comment on rumour. These are pieces we found thoughtful enough to link to.
"The Geneva startup making lawyers admit AI is no longer a parlour trick."
"What ZAAN gets right is what nobody else has bothered to: the audit trail."
"The Open Bar program is the most interesting thing happening in legaltech right now."
Find us in Geneva, Shenzhen, or your inbox.
If you're a customer, a candidate, a journalist or a clinic — write. We read everything that arrives, and reply to most of it.