AI Built for Law: Thomson Reuters CPO David Wong on why purpose-built tools matter in legal

AI is really good at coming up with plans. It's not always great at executing them. David Wong thinks that gap is the whole story for legal AI right now, and it's why Thomson Reuters has been building the way they have. David is Chief Product Officer at Thomson Reuters, which has been around for over a hundred years and is one of the most established names in legal information. He leads CoCounsel, their AI platform, which recently hit 1 million users across 107 countries. He also just published a widely read Fortune piece laying out his framework for where legal AI is heading. We recorded the day after Anthropic launched Claude for the Legal Industry, with both Thomson Reuters and Descrybe included as MCP connectors. So this is a conversation about what it looks like when a 100-year-old company and a three-year-old startup end up in the same lineup, and what separates purpose-built legal AI from generic tools. We dig into: • The plan vs. execution gap at the center of David's Fortune piece, and why generic AI tools often stumble at the second part • The "wrong tool for the problem" framing, and why hallucination isn't the real risk; choosing the wrong tool is • Why both Descrybe and Thomson Reuters count as purpose-built tools, even though they sit at very different ends of the legal tech market • Thomson Reuters's "we replaced ourselves" moment: rebuilding Westlaw on a generative AI foundation, and why staying still wasn't an option • The PC software ecosystem analogy David uses to explain where this is all heading, and why no single AI company is going to write every legal application • Claude as orchestrator, not soloist: what it looks like when frontier AI delegates to a "team of specialists" through MCP connectors • The access to justice payoff both ends of the legal tech market are betting on David has been at the front of Thomson Reuters's AI strategy from the beginning, and his Fortune piece moved a chunk of the current debate about how AI actually executes on legal work. But this conversation stays grounded in customers, coexistence, and what it actually takes to ship this stuff responsibly at scale.