Legora is building the agentic operating system for legal work

Legora, a Stockholm startup founded in 2023, wants to be the platform underneath legal work, and OpenAI is spotlighting it as a showcase customer. The numbers it cites are big: its own legal team says NDA review fell from about a fifth of its capacity to under 1%, and one Fortune 100 customer reportedly cut contract turnaround by 90%, saved over $20 million on outside litigation and diligence, and got back 9.5 hours per lawyer each week.

Law is a profession where time is billed and errors are expensive. If AI can broaden review without breaking trust, that changes how legal services are priced and staffed.

Let's be clear about the source first. This is an OpenAI customer story about a company that runs on OpenAI models. The figures come from Legora's own executives and an unnamed customer. That does not make them false, but it does make them marketing, and we would like to see them hold up under independent scrutiny.

With that said, the idea we find most persuasive here is breadth, not speed. In traditional due diligence, cost pressure forces lawyers to go deep on the most important contracts and skim or skip the rest. A tool that can run the same questions across thousands of agreements means risks hiding in the long tail can at least get flagged for a human. That is AI expanding the scope of careful work rather than replacing judgment, and it is the version of legal AI we want to see win.

The NDA playbook is also a good template: the team studied what it had historically signed, encoded its own limits, and let the system take a first pass while lawyers handle the exceptions. Boring documents are exactly where automation should land first.

The risks are real, though. Legal mistakes do not announce themselves; a missed clause shows up months later in a dispute. When review becomes nearly free, the temptation is to trust the first pass and let attention thin out. And there is a pipeline question we keep returning to: if junior lawyers no longer grind through stacks of contracts, where do future senior lawyers learn the craft? The company's CTO talks about overnight diligence over millions of pages. Impressive, and also a lot of output for a human to meaningfully check by morning.

We are bullish on the direction. We will judge it on error rates, accountability when it misses something, and whether firms keep training the people who are supposed to supervise it.

Legora says OpenAI models power core workflows including agentic legal work and large-scale document review, and that its teams also use ChatGPT, Codex, and other OpenAI tools internally. It reports growing from three offices to 18 in the past year, along with three financing rounds and five acquisitions. A Legora plugin also lets legal teams use its capabilities inside ChatGPT.