Legora cut NDA review from a fifth of its legal team's time to under 1%. The figures come from OpenAI, not an auditor

On October 28, OpenAI published a customer story about Legora, a Stockholm company that sells AI to lawyers. The headline figures look impressive. Nobody has verified them independently, though, and the key customer stays anonymous.

Legora's evidence is mostly its own legal team and one unnamed client

Legora was founded in Stockholm in 2023. CEO Max Junestrand calls it “the agentic operating system for legal work”. According to the piece, OpenAI is a key model provider for the platform's core workflows, from agentic legal work to large-scale document review.

The most concrete example comes from inside the company. General Counsel Emma Kadri Bergström says reviewing incoming NDAs used to take about 20% of her team's capacity. Now it is under 1%. The team built a playbook from agreements it had signed before. The system takes the first pass and flags terms that fall outside acceptable limits.

The second example is a Fortune 100 customer the story does not name. According to Junestrand, it cut contract turnaround time by 90%, saved more than $20 million on external litigation and diligence, and gave each lawyer back 9.5 hours a week. CTO Jacob Lauritzen adds that coding agents let the team explore 20 solutions at once. Going from bug report to working fix sometimes takes around five minutes.

The story lands in the week Legora needs a big vendor's stamp

The timing is worth noting. On October 21, Legora announced a $550 million Series E at an $8.8 billion valuation, up from $5.6 billion in its previous round. Over the past year, the story says, it grew from three offices to 18, closed three funding rounds and made five acquisitions.

This is OpenAI's second piece on Legora in two months. In September it described Legora using GPT-6 Astra to review 41 documents and catch all four planted errors. Law is a rewarding shop window for model vendors: expensive hours, mountains of text and clients eager to hear about savings. On the same day the Legora story ran, Anthropic added ready-made artifacts for legal teams to Claude for Legal. The model makers themselves are competing for lawyers.

One more detail matters for legal departments. The story calls OpenAI a “key” provider and says nothing about exclusivity. What Legora sells is mainly the layer above the models: playbooks, review and workflows. Buyers pay for that layer, and the model underneath may change over time.

The 90% saving belongs to one client nobody can name

Every figure comes from Legora and appears on OpenAI's website. There is no methodology. We do not know how turnaround time was measured, what baseline the $20 million saving uses or how many contracts went through review. The NDA numbers describe Legora's own team, the most motivated user its product will ever have.

The story also concedes that lawyers still review and interpret what the agents produce. In due diligence, that moves the bottleneck. When the system reads thousands of contracts instead of hundreds, someone has to assess every risk it surfaces. The piece does not say how many flagged risks turned out to be real.

Named clients and counts of missed clauses will settle it

Legora says more than 140,000 legal professionals at over 2,400 organizations use its platform. Whether the story's numbers hold up will become clear only when named clients publish their own measurements. Above all, we need error data: how many problematic clauses the system missed and how many false alarms a lawyer had to throw away.

The second signal is price. If AI really returns 9.5 hours a week per lawyer, firms that bill by the hour will have to explain to clients what they are paying for. That is where we will see who keeps the savings.

Lilith's verdict

Legora had OpenAI issue its report card and copied the straight A's from its own notebook. When a law firm sends a client an invoice 9.5 hours a week shorter, we will know who the savings really belong to.