Thomson Reuters reports CoCounsel has 1 million users, shares soar

Hi Andrew, thanks for having me.

No, it was developed a number of years ago, Andrew. What we've done is we've created an AI platform for fiduciary professionals, so as you said, lawyers, tax and accounting professionals, auditors. Essentially, we developed it a couple of years ago and we've developed it in such a way that it is model agnostic. So in other words, we can take the very best enterprise-focused foundation model and swap that in. Now, at the moment, in our view, based on our testing, that's Anthropic's Claude suite of models. We have, though, developed our own model called Thompson, which is starting to outperform many of the models specifically for legal. And so it wasn't developed with Anthropic or Claude, but certainly we think their enterprise models are the best at the moment. But we've created some flexibility as we go forward.

Yeah, I mean the legal profession sort of in its current form I think is probably 300 to 350 years old. You know, it was sort of started in the tea houses in the city of London all those centuries ago and it sort of fundamentally hasn't changed all that much. Certainly word processing technology and e-discovery and other things have made some differences, but in our view, this will enable lawyers to be very, this being AI, will enable lawyers to be very, very significantly more productive. And I think what investors missed with the sort of Anthropic announcements and some of the moves of the foundation models is at Thomson Reuters we serve fiduciary professionals. And I think we have to distinguish, and investors, you know, I don't think have done this yet, between sort of knowledge workers, so that's general office workers at one level, professionals, so whether that's salespeople, communications, marketing, architects, whatever it might be, from fiduciaries. Fiduciaries being lawyers, tax, accounting, audit and the like have to be right. I mean the consequences of being wrong are very, very severe for fiduciaries. And what we do is we combine a model like Claude with, I think most importantly, our content. And we've got centuries of curated, highly specialized, very accurate content with deep domain experts, thousands of them, who train our agents. And what it does is it produces results that are fiduciary grade versus results that may be 95% right, which is not acceptable to these fiduciary professionals.

I think they are. I mean, I think they missed this point about the quality and the importance of content and expertise to train these agents in this environment. We think we're the only player who can do that effectively and meet or exceed the needs of these fiduciaries. And so we think the investors have missed it certainly over the last six or seven months and particularly over the last few weeks.