Robert Ambrogi Blog
CoCounsel, which launched almost exactly three years ago, on March 1, 2023, as the first AI legal assistant built on GPT-4, today marked a notable milestone, reaching 1 million customers across 107 countries and territories.
Developed by legal research startup Casetext, Thomson Reuters acquired CoCounsel (and Casetext) just four months after its release, for a whopping $650 million in cash. Since then, TR has expanded CoCounsel across its product lines and across professional verticals, from legal to risk, compliance, tax, accounting, audit and trade.
Thomson Reuters (TSX/Nasdaq: TRI) announced the milestone at an event in San Francisco Monday, framing it not just as a measure of growth but as a signal of a broader shift in how regulated industries are approaching artificial intelligence — from cautious pilots and experiments to embedded production systems.
“Professionals are not deciding whether to use AI anymore,” said Steve Hasker, TR’s president and chief executive officer. “They are deciding which AI they trust when their reputation and their clients’ data are on the line.”
The milestone comes as AI adoption among legal and other regulated professionals has moved decisively past the experimentation phase, TR says. Firms and corporations are no longer asking whether AI belongs in high-stakes workflows — they are choosing which systems can actually meet the standards those workflows demand.
AI for Regulated Work
While CoCounsel started as a standalone chat interface, TRThe company has emphasized that general-purpose AI, however capable, falls short in professional environments where outputs must withstand courtroom scrutiny, regulatory review or audit proceedings. CoCounsel is designed specifically for those contexts, TR says, grounding its outputs in editorially refined legal and tax content developed over 175 years and validated by more than 4,500 TR subject-matter experts across legal, tax and compliance domains.
While CoCounsel started as a standalone chat interface, TR has integrated it across its major product lines, including Westlaw and Practical Law on the legal side and Checkpoint on the tax side, as well as Microsoft 365. It draws on a multi-model AI architecture — working with frontier models from Anthropic, OpenAI and Google, alongside TR’s own proprietary AI technology — and produces structured, citation-backed outputs rather than freestanding text generation.
On data privacy, TR has stressed that customer inputs are never used to train third-party models or generate outputs for other users.
“One million professionals have chosen CoCounsel,” David Wong, TR’s chief product officer, wrote in a related blog post. “Not for pilots. Not for experiments. As core infrastructure for how they work.”
The Road to a Million
CoCounsel’s path from launch to a million users involved considerable iteration. In a separate blog post, Joel Hron, chief technology officer at TR, described the early months after Thomson Reuters introduced AI Assisted Research — the first generative AI feature in Westlaw — as challenging and formative.
“What felt strong in our research loops needed refinement when put to the test with real human feedback,” Hron wrote. Over time, he acknowledged, an intense focus on accuracy created its own trade-off, in that the system became highly reliable but less fluid. “We optimized for never being wrong. Our users wanted us to also optimize for being genuinely helpful.”
That tension, Hron wrote, drove CoCounsel’s evolution into something more ambitious — including the development of Westlaw Deep Research, which the company describes as the most advanced AI-powered legal research system available, capable of analyzing thousands of documents, synthesizing findings across jurisdictions, and delivering court-ready analysis with citations and reasoning.
What Comes Next
TR used the million-user announcement to tease what it said will be the next generation of CoCounsel Legal, which is entering beta soon.
The new version is designed around conversational task execution — allowing a lawyer or legal professional to describe an objective much as they would brief a colleague, and then have CoCounsel build a plan, retrieve authoritative sources from Westlaw and Practical Law, search relevant documents and precedent, verify citations, and deliver structured work product within a single system.
Additional next-generation capabilities within CoCounsel Tax and ONESOURCE+ are planned for later in 2026.
The company has also disclosed it is developing a proprietary large language model designed specifically for legal, tax and compliance use cases — a move that would reduce dependence on external model providers for its most sensitive professional applications.
Hron, in his blog post, acknowledged the ambition of what lies ahead while expressing confidence in the team that got CoCounsel to this point. “One million users proves we are trusted. What we’ll prove when we 10x that number this year is that we’ve built the AI professionals genuinely can’t live without.”
Thomson Reuters invested more than $200 million annually in productized AI, the company has said, and has indicated it has approximately $11 billion in capital capacity through 2028 for continued investment and acquisitions.