Thomson Reuters CEO Steve Hasker on the Next Generation of CoCounsel, the ‘Future of Professionals’ Report, and Why TR Is Building Its Own LLM | LawSites

Thomson Reuters says its new CoCounsel proved itself in the best beta trial its CEO has ever seen — but it won't say by what measure

Thomson Reuters has opened early access to a rebuilt version of CoCounsel Legal, which the company describes as the biggest overhaul of its AI assistant since it bought Casetext in 2023. The release landed alongside TR's Future of Professionals Report 2026, which itself warns of a growing gap between how much organizations are adopting AI and how much value they're actually getting from it. In an interview with LawSites' Bob Ambrogi, CEO Steve Hasker said the new CoCounsel finally puts Westlaw and Practical Law content, plus the training input of TR's 2,700 attorney-editors, to work inside the agent — something the prior version didn't do. He also confirmed the product runs on Anthropic's Claude models via the Claude SDK, that TR treats itself as "model-agnostic," and that the company is separately building its own LLM, referred to in the piece as Thomson.

What actually changed in CoCounsel Legal?

TR rebuilt CoCounsel on the same architecture as Westlaw Advantage, finally connecting its licensed content and editorial staff to the agent itself.

That's a real architectural claim, not a cosmetic one — the earlier CoCounsel, built in the Casetext-acquisition era, apparently didn't route its editorial content and attorney expertise into the model the way the new version does. If true, it closes a gap that competitors have pointed to for a while: a legal AI tool is only as good as what it's grounded in, and TR is asserting it's now grounding CoCounsel in the same assets that make Westlaw defensible.

How does TR back up its confidence in the results?

Mostly with Hasker's own testimony — he calls it the best beta he's personally seen, without publishing accuracy or reliability figures.

Hasker told Ambrogi this is "the most successful beta trial I've ever been a part of, by a long way, in terms of the feedback on the utility of the product, the accuracy of the product, and also the reliability." That's a CEO's characterization of customer feedback, not a published benchmark, a sample size, or an independent test. The Future of Professionals Report published the same day is explicitly about the disconnect between adoption enthusiasm and measured value — which makes it notable that the CoCounsel news arrives without any measured value of its own attached.

Why is TR building its own LLM if it calls itself model-agnostic?

TR is running Claude in production today while developing a proprietary model in parallel, hedging against being tied to any one provider.

Hasker was direct that TR picked Claude models for CoCounsel and Westlaw Advantage based on ongoing weekly evals, and that the relationship with Anthropic's engineering team has been productive. Building "Thomson" alongside that isn't necessarily contradictory — plenty of platforms run a licensed model now while keeping optionality for later — but buyers should understand that the model underneath their workflow could change, and ask what happens to accuracy, cost, and data handling when it does.

What should e-discovery and litigation-tech buyers ask before trusting this?

Ask for the actual beta numbers TR isn't publishing, and how the vendor defines "accuracy" and "reliability" for legal work.

For a reader evaluating whether to trust an AI system with privileged evidence, "successful beta" is not a specification. Ask what accuracy was measured against, whether results were reviewed by attorneys outside TR, and whether outputs cite back to source documents the way a defensible record requires. Ask, too, whether your evidence trains any underlying model — TR hasn't detailed that here, and it's the question that matters most before privileged material touches any AI tool.

Frequently asked questions

Is the new CoCounsel generally available?

No — Thomson Reuters has opened early access only, with customers who joined the beta trial ahead of a wider rollout.

Does this replace Westlaw Advantage?

No. Hasker described CoCounsel as built on the same underlying architecture and team as Westlaw Advantage, but as a separate flagship product.

Did Thomson Reuters publish benchmark data for the new CoCounsel?

Not in this interview. The claims of high accuracy and reliability come from Hasker's characterization of beta feedback, not published figures.

Sources: LawSites / LawNext, "Thomson Reuters CEO Steve Hasker on the Next Generation of CoCounsel, the 'Future of Professionals' Report, and Why TR Is Building Its Own LLM," by Bob Ambrogi.