Last August at ILTACON, Thomson Reuters CEO Steve Hasker told me about the company’s decision to acquire a bunch of genius scientists building a smaller, more manageable legal model. It stood out because everyone crammed into that exhibit hall couldn’t wait to explain how embedded they were with OpenAI or Anthropic. And Thomson Reuters was — to be fair — bragging about their frontier connections too. But it struck me that Hasker saw then a benefit to building out some potential independence if things went sideways in AI.
By then I’d already been talking about the importance of an “American DeepSeek” — a shorthand for a more economical model that may not solve the OpenAI delusion of a mechanical messiah, but would deliver solid productivity gains while also being able to admit what happened in Tiananmen Square. So it intrigued me that any major company in 2025 had invested in a backstop in case a foundational model choked on its own valuation.
Now, Thomson Reuters has launched Thomson 1.0, built by that team of experts. The company also just rebuilt CoCounsel from the ground up. And Greenberg Traurig deployed the whole next generation of CoCounsel Legal across all of its almost comically numerous offices, capping years of the firm’s lawyers beta testing the project and helping Thomson Reuters get it all together.
It’s hard to think of a better partner for this sort of endeavor than Greenberg Traurig. The firm has over 3,000 attorneys and practicing in New York vs. Tallahassee vs. Riyadh couldn’t be more different. Not that every office deployed the technology the same way, but Thomson Reuters partnered with a firm that thinks about what it means to practice in locations as diverse as New York, or an anti-science, repressive regime, or Saudi Arabia.
The announcement keys off one of the big themes of this ILTACON: “what if we all just play along?” If the last couple years were defined by catchphrases like “single-pane,” imagining a world where one provider would control the entry point to every lawyer’s working day, this year is all about opening the door to meet lawyers where they are. Some lawyers will open their day in this platform… some in this other one… on that front, it’s a competition for eyeballs. But at the end of the day, they all need access to key context, and that’s where a company like TR leans on the advantages it knows it has.
The Tokengeddon is coming. Ironically, as legal faces pressure to step back from the billable hour to keep charging proper value for a service that takes less and less time, the AI universe is eventually going to move away from flat fees to consumption pricing out of necessity. And when that happens, firms are going to have to be ready to get actionable work in a cost-effective manner.
The next generation of CoCounsel Legal helps [Greenberg Traurig’s] lawyers turn insight into action by combining trusted legal content, agentic intelligence, and workflow execution in a single experience, grounded in the transparency and accountability of our Fiduciary-Grade AI™ standard.
Digging into what that means in practice with Matthew Beekhuizen of Greenberg Traurig, it struck me that — after months of AI companies preaching “agentic” as a never-ending automation train that leaves a lawyer to sort it all out on the back end — the Thomson Reuters workflow that Greenberg Traurig’s been working with is designed to give the lawyer easy opportunities to jump into the process and check in. I’ve mused that “human-in-the-loop” doesn’t mean anything if the human comes in at the end and is invited to scrounge around through a giant log of agent actions. That would be “auditable,” but it’s not a realistic prospect. Beekhuizen described the CoCounsel interface as psychologically inviting of lawyer involvement.
Hasker added that the CoCounsel skills will take drafting briefs down from 20 or 30 hours down to two or three. But — and this is the critical part — not down to two or three minutes. Theoretically, one could build something that speedruns to a final draft, but that draft is going to suffer. There’s a sweet spot that allows the human to remain engaged, to check as the process develops, and to engage in real-time to guide the agentic process along the path.
Human-in-the-loop always read like an afterthought. You tell someone you’re keeping them “in the loop” when you’re about to forget to invite them out. Alas, it became the industry catchphrase and we all had to live with it. But the human needs to be centered — in the thick of the process as it develops. The trick isn’t keeping the lawyer in the loop, but making the loop somewhere a lawyer would want to be.
Earlier: Dispatches From The AI Bubble: ILTACON 2025
Legal AI And The ‘New And Improved Detergent’ Problem
Thomson Reuters Rebuilt CoCounsel: A Pivot
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news.
The post Greenberg Traurig And Thomson Reuters Tackle What ‘Human In The Loop’ Was Missing appeared first on Above the Law.