TLDR
- Theo AI builds AI models that predict likely settlement values, helping corporate legal teams decide whether to fight or settle cases.
- The startup has raised more than $10 million and launched a General Counsel Advisory Board with leaders from Bristol Myers Squibb, Docusign, GoDaddy, HP and others.
- Unlike contract-review tools, Theo AI focuses on the ‘data layer’ of litigation, using private settlement records and multimodal analysis to benchmark outcomes at scale.
Theo AI is trying to solve one of the most opaque problems in corporate litigation: determining how much a legal case is likely to settle for before it ever reaches a courtroom.
The Palo Alto–based startup has built an AI platform that ingests thousands of pages of legal documents, medical files, insurance records and internal corporate data to predict the settlement amount a company should expect to pay in a dispute. It claims 85% accuracy.
It may be a surprise to know that nearly all civil cases settle out of court. “You may think judges decide outcomes for most litigation, but that’s only about 3% of cases. The vast majority – 97% – get settled out of court,” said Theo AI CEO and cofounder Patrick Ip, in an interview with The AI Innovator. “The magic number that you want to know is, outside of the court between two parties, what’s that number we need to reach so that we can both close up on that litigation?”
But because settlement amounts in private disputes are not publicly disclosed, corporate legal teams often operate with little reliable data when deciding whether to fight or settle.
Theo AI’s system benchmarks against both industrywide settlement patterns and a company’s own history – data that many legal departments do not have organized or easily accessible. “You would think with these corporates, they maybe should have a directory of all their settlements that’s nicely lined up, but oftentimes they don’t,” Ip said. “We do all the work of organizing all their previous settlements.”
The startup focuses on high-volume civil litigation found in industries such as retail, where slip-and-fall injuries and liability claims can number in the thousands each year. “You can get a very smart lawyer to look at one case and give you a pretty good assessment, but that one extremely experienced lawyer can’t scale to a thousand a month,” Ip pointed out. “That’s really where we start to come in.”
Theo AI helps lawyers benchmark, triage, and determine which cases should be litigated and which should be settled. The startup breaks cases into “agentic nodes,” specialized modules that analyze specific evidence, including medical records, insurance data or images. In one matter involving a hair-relaxer product claim, the system evaluated images to determine whether a customer had used a product at the center of the dispute.
This multimodal structure – built primarily on Google’s Gemini, fine-tuned for vertical legal tasks – helps the platform bring out facts and determine relevance without requiring users to label or manually categorize documents. “We can passively take everything, ingest it, and now you have all this great data to help you make the right decisions,” Ip said.
85% accuracy rate
Theo began in 2024 by testing its first models with litigation funders, which act as financial backers for small plaintiffs in exchange for a cut of legal wins. These funders often have access to private settlement data. When funders ran historical cases through Theo AI’s system, the model correctly predicted outcomes about 85% of the time – versus roughly 60% to 65% for human reviewers, according to Ip.
The startup has since shifted focus from litigation funders to large corporate legal departments, especially those that deal with thousands of cases per year. Ip said Fortune 500 chief legal officers have joined Theo’s advisory efforts because they want more empirical guidance when managing litigation risk.
That effort became formal with Theo AI’s new General Counsel Advisory Board. The board includes senior legal leaders from Bristol Myers Squibb, DocuSign, GoDaddy, HP, SentinelOne, U.S. Bank and other major companies. The group gives corporate legal executives visibility into how Theo’s predictions are built and validated, an important factor in adopting AI for high-stakes decisions.
“The work of Theo AI has the potential to revolutionize high-stakes litigation settlements,” said Sandra Leung, former general counsel at Bristol Myers Squibb, in the press release. “It will expedite settlements and reduce legal expenses.”
Jessica Nguyen, deputy general counsel for AI Innovation and Trust at DocuSign, added that the model replaces guesswork with measurable data. “Theo transforms a theoretical risk to a number grounded in data,” she said.
Alongside the new advisory board, Theo closed a new funding round led by Run Ventures, bringing its total raised to more than $10 million. The firm’s partner, PT Ungvichian – formerly on the founding team of SoftBank’s Vision Fund – will join Theo’s board of directors. “They’re tackling one of litigation’s biggest hurdles, uncertainty,” Ungvichian said in a statement, adding that the firm views verticalized legal AI as an emerging category.
The funding follows a $4.2 million seed round raised in May and will help Theo expand its engineering team and extend its models into additional categories including employment disputes and later commercial and patent litigation, where exposure can reach into the billions. Today, the startup employs 11 people and operates through Stanford University’s StartX accelerator ecosystem.
While Theo AI does not automate legal drafting or contract workflows – a crowded area led by startups such as Harvey, Spellbook and Robin AI – it aims to own the ‘data layer’ of litigation, acquiring unique settlement information that larger incumbents such as Thomson Reuters cannot access.
Ip said he eventually wants the technology to reach small businesses and consumers, though today the focus remains on Fortune 500 legal teams.
