Legora brings OpenAI models into legal workflows and its own teams

OpenAI said Legora is using its models across core legal workflows, from agentic work to large-scale document review. Legora is also expanding ChatGPT Work and Codex across its own teams. The customer profile identifies the products and workflow categories involved, but not the model versions behind the broader production deployment. It also provides no seat counts, adoption rates or usage logs for the internal rollout.

Legora says its Agent product plans and carries out multistep work across documents, data sources and tools, checks its progress, and presents the result for human review. Its Tabular Review product turns imported document sets into a spreadsheet-like table, with access to the source document behind each cell. The pages establish the workflow design Legora advertises, not the system’s accuracy or output quality. The deployment follows OpenAI’s Agreement Analyzer work with Sullivan & Cromwell in another legal-review setting.

OpenAI also lists a Legora plugin for ChatGPT developed by Legora AB. According to the listing, ChatGPT can send jurisdiction-scoped legal questions to Legora and return grounded answers with links for users to inspect during citation review.

A separate customer statement corroborates some of those use cases. Bryan Cave Leighton Paisner said it selected Legora after a multimonth pilot involving more than 100 lawyers and spanning document review, due diligence and legal drafting. That confirms Legora’s use in those workflow categories, although the law firm’s qualitative assessments of efficiency and accuracy remain participant-reported.

The numerical results in OpenAI’s new profile come from Legora executives and have not been independently verified. Chief Technology Officer Jacob Lauritzen said coding agents let teams explore 20 solutions at once and can sometimes cut the time from a bug report to a working fix to about five minutes. The profile gives no benchmark design, sample size or external validation for those figures.

Legora General Counsel Emma Kadri Bergström estimated that a playbook-driven first pass reduced NDA review from about 20% of her legal team’s capacity to less than 1%. OpenAI’s profile does not explain how the capacity was measured or specify the measurement period.

The biggest claims involve an unnamed Fortune 100 customer. Legora CEO Max Junestrand attributed a 90% reduction in contract turnaround time, more than $20 million in savings on external litigation and diligence, and 9.5 hours returned to each lawyer per week to that customer. The profile does not disclose the customer, baselines, measurement period or methodology, preventing customer-side confirmation. It also says organizations where more than 75% of leaders actively use Legora have three times higher median usage, without identifying the sample, definitions, period or a causal analysis.

OpenAI published a separate Legora case study on Sept. 3 about a financial-statement tie-out across 41 documents in one Agent run using GPT-6 Astra, with final judgment left to legal professionals. Legora said the run finished in minutes, found all four planted errors and improved nearly 40% over the previous model on that workflow. Legora designed and reported the test without publishing the underlying results, so those performance figures also remain unverified company claims.