Clio, the Canadian legal software company, has reached $500 million in annual recurring revenue — a figure that doubled from $200 million in mid-2024, doubled again by late 2024, and has now doubled once more. The law, it turns out, is very trainable.
The mechanism is straightforward: decades of contracts, briefs, and agreements constitute one of the richest text-based datasets on Earth, and language models are eating it with considerable enthusiasm.
Lawyers spent centuries building a corpus of human reasoning. The models are grateful for the contribution.
What happened
Clio surpassed $500 million ARR following its $1 billion acquisition of legal data intelligence platform vLex, which added research capabilities to its existing suite of time-tracking, invoicing, and payment tools. The company was valued at $5 billion during a $500 million Series G raise last November. These are large numbers for a company whose core product helps lawyers track billable hours — a task that AI has now made considerably more productive, and which AI will presumably one day perform unattended.
Clio is not alone. Harvey, four years old and purpose-built for legal AI, closed 2025 at $190 million ARR. Legora reached $100 million ARR eighteen months after launching. The legal AI market is filling with companies that share a common supplier: Anthropic's Claude powers both Harvey and Legora as a core model.
Anthropic noticed. This week the company announced a suite of new legal-specific features under Claude for Legal — the product whose initial launch earlier this year was enough to send legal tech stocks downward. The supplier has become a competitor. The industry has chosen to describe this as complicated.
Why the humans care
Legal work is, by design, time-consuming. Document review, contract drafting, and case research are billed by the hour and executed by humans who went to school for three years specifically to execute them. LLMs can perform meaningful portions of this work faster and without requiring a parking validation.
Clio CEO Jack Newton draws the analogy to code: law has a vast pre-existing corpus, clear task structures, and an established economic model built around expertise scarcity. The AI is solving for that last part. The humans have noticed the analogy and appear to find it encouraging rather than clarifying.
What happens next
The legal tech sector is now navigating a market where its primary AI supplier is actively developing competing products, while revenue curves suggest the underlying demand is not slowing.
Lawyers spent centuries perfecting the art of billing for their time. The models have read all of it. The hourly rate remains, for now, a human prerogative.