Loading page…
An AI assistant built for lawyers, sold to Thomson Reuters for $650M in cash
$20M revenue and 100 people at the time of the pivot
The crux is evaluation and breaking a professional workflow into steps — most people ship a flashy 60-70%-accurate demo and go raise money; the ones who actually win spend two full weeks grinding one prompt from 61% accuracy up to 97%.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Founder Jake is a lawyer by training who started CaseText in 2013 building legal research tools. Back then the business wasn't sexy — lawyers collectively earn in the trillions, but they spend pitifully little on software, so before LLMs arrived legal tech had a very low market ceiling. Jake later admitted that when he picked this vertical he underestimated how small the market was, and that was a mistake. The turning point came in the summer of 2022. The company was doing about $20M in revenue with close to 100 people — a mid-size SaaS living comfortably but hardly exploding. That's when they got early access to GPT-4, a window only a handful of companies had. Jake then made a call most people couldn't: shut down the entire existing, revenue-generating business and go all in on rebuilding a brand-new product, CoCounsel. CoCounsel is an AI assistant for lawyers: it reads documents, runs legal research, reviews contracts, and organizes evidence. Once the product got good enough, word-of-mouth referrals and press coverage came to them, and the existing sales team essentially became order-takers. About two years ago, legal information giant Thomson Reuters bought it for $650M in cash. That's a public acquisition event, not a valuation story — real money in the bank.
CoCounsel launched into a base CaseText already had, roughly 100 people and $20M of revenue built since 2013, but the product itself was new and lawyers are conservative buyers. Two moves did the convincing.
The quick-reference card — acquisition channels, replication playbook, and risk map — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself.
The pricing decision is the whole lesson. Comparable legal tools sold at $20 a month, while CoCounsel charged around $500 per contract or $6,000 per seat per year, because the thing being displaced is a paralegal's hours rather than a software seat, and at that rate the customer is still paying far less than a salary.
Accuracy is the flywheel, and the threshold is specific: the point where a lawyer stops re-checking the output. Below it, every answer costs the customer verification time and the product is a curiosity; above it, the work is actually handed over and the tool becomes load-bearing.
This business ran two long stretches without external signal, and handled them in opposite ways. The first lasted nine years, from founding in 2013 to the summer of 2022, as a legal research company in a market where lawyers bill trillions and buy almost no software.
The rest of this breakdown — 10 more sections — is in the Unicorn Club, with every section of all 1,335 ideas: first customers, first dollar, the flywheel, and the silent stretch.
Join the Unicorn Club — $5/moCancel anytime, in one click