Legora’s Jacob Lauritzen explains building legal agents without seeing client data
Legora’s CTO, Jacob Lauritzen, discusses how the legal AI company develops agents that handle sensitive client data without direct visibility. Founded in a Stockholm basement with ChatGPT and Swedish law tools, Legora now offers features like drafting documents, reviewing 10,000 contracts at once, and coordinating workflows between lawyers, clients, and other agents. The company uses simulated…
Key points
- Legora tests agents in simulated environments modeled on real customer workflows, improving accuracy without direct data access
- Citation accuracy in drafts improved from 70% to 99% after one release, showing iterative testing’s impact
- Legora runs on Azure for privacy, avoids open-weight models, and now employs 100 engineers after starting with 3
Legora runs everything on Azure for privacy and security, avoiding open-weight models and ElevenLabs. The company’s Tabular Review tool—a million-cell spreadsheet—has become an industry standard. Legora’s 100 engineers (up from 20) now struggle to keep pace with demand, despite Lauritzen’s early bet that 20 engineers would suffice. The firm also uses its own product for acquisition due diligence and is hiring engineers in New York, Stockholm, and London. Lauritzen speculates that 60% of legal work is already handled well by models, but the remaining 40% requires smarter, more specialized agents.
Building Legal Agents When You Can't See the Data - Jacob Lauritzen, CTO of Legora
Latent Space · 2 November 2026
Loading the full article…
This text was published by Latent Space and written by Latent.Space. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- OpenAI releases LifeVal evaluation for AI task performance · 1 src
- Nvidia and AMD back $6B AI startup Hark to challenge Grok · 1 src
- Sierra integrates Ghostwriter into Agent Studio workflows · 1 src
- Claude Code deletes local coding sessions after 30 days · 1 src
- Anthropic adds message threads to Claude Platform for faster agent responses · 1 src
Comments
via GitHub Discussions