Legora Uses OpenAI to Reshape Legal Work: 90% of Contracts No Longer Slip Through Review

The legal industry has long been seen as one of the slowest sectors to adopt AI, but Legora is breaking that stereotype. Founded by a college dropout, this legal AI platform embeds the GPT series models into every part of a lawyer's daily work, turning contract review from "only picking the most expensive 10%" into "full-coverage, second-scale screening." More notably, Legora's founder says outright: without OpenAI, this company would not exist at all.

From Dropping Out to Founding a Company: A Life Trajectory Changed by an API

According to the Legora founder's recollection in an official OpenAI video, the open access to GPT 3.5 was the direct motivation for him to drop out of school and start the company. This detail itself illustrates one thing: the availability of foundation models is reshaping the threshold structure for entrepreneurship. In the past, building a legal tech product required an in-house NLP team, annotated data, and model training, with cycles measured in years; now that powerful models are available via API, a small team that understands industry pain points can quickly get a product off the ground.

Legora's positioning is not complicated—enable lawyers to conveniently use AI in every task throughout the day. This "embedded" approach differs from making AI a standalone tool; it is more like folding model capabilities into existing workflows to reduce friction for lawyers.

Open access to GPT 3.5 refers to OpenAI making the gpt-3.5-turbo model available to developers via API in March 2023. Unlike the earlier GPT-3 API, which required an application queue and charged expensive per-token fees, gpt-3.5-turbo drastically reduced costs (about $0.002 per thousand tokens), responded faster, and supported conversational interaction. This move is regarded as the starting point of the large-model application boom: developers no longer need to build their own compute clusters or train models—they can simply call an API to embed powerful language capabilities into their own products. For vertical fields like legal tech, this means startup teams can focus resources on industry understanding and workflow design rather than underlying AI technology R&D, greatly lowering the threshold and timeline for productization.

Finally, Someone Is Looking at the 90% of Contracts

In traditional legal practice there is an open secret: due to time and labor cost constraints, teams often conduct in-depth review only on the 10% of contracts with the highest value, while the remaining 90% are basically left to "spot checks." This is not because lawyers are irresponsible, but because it is not economically worthwhile.

What Legora changes is this cost curve. The founder puts it directly: the 10% can still receive in-depth review, but the other 90% can now be screened by AI "almost instantly." This means an order-of-magnitude expansion in risk coverage—long-tail contracts that were previously abandoned by default can now enter the review scope.

Regarding concrete throughput, the founder gave a scenario: 10,000 contracts, 10 queries per week, covering all contracts, completed in about a few minutes. This scale is almost unimaginable in manual mode, and it is precisely the fundamental change AI brings in the dimension of "breadth."

"Full coverage" and "sampling review" in contract review are two key concepts in legal risk management. Traditionally, due to limited manpower, legal teams often adopt a value-based sampling method: only contracts with the highest amounts or greatest risks are reviewed clause by clause, while the rest receive only formal checks or none at all. The hidden danger is that long-tail contracts may contain clauses deviating from standard terms (such as unreasonable liability caps or automatic renewal traps). AI-driven full review uses natural language processing to quickly compare contract clauses against a predefined rule base, identify anomalous clauses, and flag risk levels, making review tasks that were previously abandoned due to cost economically feasible and thereby raising the overall compliance level.

Agent Workflows: Due Diligence Goes from Manual Grunt Work to "Launch and Run"

What excites the founder most is the end-to-end AI due diligence workflow. The logic is: a lawyer launches a task, the agent automatically carries out contract review and other types of work, and the results are handed to humans for review.

This marks a re-division of roles—AI takes on massive, repetitive, pattern-based initial screening labor, while humans retain judgment, decision-making, and responsibility. For the legal industry, this division of labor is closer to reality than the narrative of "AI replacing lawyers," and is easier for professionals to accept.

AI Agent refers to an AI system capable of perceiving its environment, planning autonomously, and executing multi-step tasks. Unlike simple question-answering models, an Agent can call tools (such as search, file reading, and API calls), decompose complex goals, execute subtasks in sequence, and adjust strategies when encountering obstacles. In legal due diligence scenarios, an Agent can automatically traverse thousands of documents in a data room, extract key clauses, identify missing information, and generate review reports, while human lawyers only need to set task goals and review the final results. This collaboration model of "humans set goals, AI runs the process" frees lawyers from repetitive labor and lets them focus on areas requiring professional judgment.

Three Rounds of Financing and Five Acquisitions in a Year: The Small Team's "Secret Weapon"

Legora's own growth data is also noteworthy. According to the founder, over the past 12 months the company expanded from 3 offices to 18, completed three rounds of financing, and completed five acquisitions in that year alone. Supporting this pace of expansion with a very small team relies precisely on its own product.

The founder's metaphor is vivid: Legora is a very fast F1 race car, and the new models OpenAI keeps releasing are engines that can be swapped in—each iteration makes the car run faster. This analogy reveals the symbiotic relationship between AI-native companies and foundation model vendors—competitiveness at the application layer partly comes from whether a company can be the first to translate new model capabilities into product advantages.

Beyond Speed, "Breadth" Matters More

The founder summarizes the change in two layers: first, the execution speed of existing tasks; second, the breadth of capabilities coverage. Speed is easy to understand, but breadth is the structural change—tasks previously excluded due to cost now re-enter the executable list. Those things "you couldn't reach before, or wished you could spend more time on" now have a solution.

His judgment is quite radical: now is the best era in the history of practicing law. If this path is right, the depth of technology's empowerment of the industry and individual capabilities will make all previous innovations seem insignificant. This carries the characteristic optimism of an entrepreneur, but the trend behind it—a sharp decline in the marginal cost of professional services and a significant amplification of individual productivity—is indeed happening.

For the legal tech sector, Legora's case provides a clear sample: foundation models supply capabilities, vertical applications handle scenario translation, and only by combining the two can "AI is powerful" be turned into "lawyers are actually using it."

One-line Summary

Legora uses GPT models to achieve full contract review and Agent-based due diligence, greatly lowering the barrier to legal AI entrepreneurship.

Paragraph Summary

Legora is a legal AI platform founded by a college dropout that embeds GPT series models into lawyers' daily workflows, transforming contract review from covering only the 10% of high-value contracts to full, second-scale screening. The founder says open access to GPT 3.5 directly prompted him to start the company, and foundation model APIs lowered the barrier to legal tech. Its Agent workflow supports end-to-end due diligence, with AI performing initial screening and humans reviewing. Over the past year the company expanded rapidly, growing from 3 to 18 offices, completing three financing rounds and five acquisitions, with its own product supporting the small team's high-speed growth. The founder believes the core change AI brings is not only speed but a structural expansion in the breadth of task coverage, and the marginal cost of professional services is falling sharply.