Thomson Reuters Monday announced what it is calling “a complete rebuild” of its CoCounsel AI tool. But more than that, it marked what President of Legal Professionals Business Ragunath Ramanathan, labeled in his presentation as a pivot to fiduciary-grade AI products. Indeed, TR gave every indication with this announcement that it may be starting to move away from an emphasis on general-purpose AI tools and toward more highly specialized products for law firms and in-house counsel handling more complex matters.

The Rebuilt CoCounsel

The new tool is designed to offer agentic services across the Thomson Reuters platforms like Westlaw and Practical Law and internal data of firms and businesses as well. According to TR, it will not only answer prompts but do things like make recommendations for next steps that perhaps need to be done. It will also be relying on Anthropic’s Claude.

It “reasons against authoritative knowledge throughout” according to the blog post by TR’s Chief Technology Officer Joel Hron, who announced it.

Here is how TR’s press release described what it can do:

Describe your matter in plain language. CoCounsel Legal creates a plan, dives deep into legal authority, reasons through the legal issues, retrieves what it needs from your own precedents, and from Westlaw and Practical Law, and drafts with citations. When new information changes the picture, it adapts. Instead of getting piecemeal answers, you get a professional standard response that you can iterate on.

I’m not sure what work will be left for us by the way.

In the LinkedIn presentation announcing the product, TR representatives called it a “total rebuild.” In the presentation and in its press release TR labeled the new tool as being able to function like a senior associate. TR representatives also talked about the enthusiastic evaluation responses received during beta and noted it will be available for early access this week.

Several TR representatives also went out of their way to reiterate that the new tools created a “fiduciary grade AI” — presumably for those who need it.

But the Announcement Seemed Different

All well and good. I am used to vendors announcing enhancements that are supposed to change the world. Often, they offer a little more than what was already out there and are designed to keep the vendors name current.

But the TR announcement felt a little different. First, instead of promoting a product that would be available sometime in the future, they announced early access would be available starting this week. That’s impressive and enhances their credibility in my book. It is also allowing attendees at the announcement access to the tool so we can play around with it. That’s a little unusual as well and suggests a certain confidence level that is often not present.

Of course, all this has to be taken with a grain of salt. We will have to see how the product performs before making an assessment of its viability and whether it’s a game changer.

TR’s Pivot

But one thing caught my eye in Hron’s blog post. His opening line of the post: “As AI becomes more capable, a new divide is emerging. Not between models, but between work that can tolerate a lack of precision and work that cannot.”

He goes on to differentiate general-purpose work that AI can do, like workflow automation, routine document processing, and open-ended brainstorming. This work is different from work that must be cited, audited, and defended. He says, “This is the work that demands something general-purpose AI was not designed for and can’t reliably deliver through prompting alone.”

This seems to mark a subtle shift in TR’s approach. With Claude, Microsoft, and OpenAI beginning to trumpet their platforms as being able to do certain legal tasks, it looks like TR is trying to differentiate certain levels of work that need that kind of validation and standard meeting assurance. It’s almost an implied admission that general-purpose AI tools do have a place for some legally related tasks. But not the more complex and sensitive matters.

It also struck me that this is a play for a market that handles more sophisticated and complex matters, (i.e., Biglaw and big business). At the briefing, they even offered a panel discussion with representatives of both to highlight the ability of the new tool to ensure confidentiality and accuracy.

TR didn’t talk pricing but I fear that the price point will be more in line with what Biglaw can and will pay than what small firms and solos can. The end result could be an increased divide between the have and have nots. You can’t blame TR for this; it’s where the money is. But that doesn’t change the end result. And it would not necessarily bode well for fundamental fairness.

Perhaps TR has seen the threat that more general-purpose providers now raise by encroaching on legal. If so, TR may be trying to double down and reiterate that their moat continues to be that they are a trusted name and that their system will allow any outputs to be defended. TR may be recognizing that the argument that public systems can’t ever be used for legal, period, isn’t what it once was, and TR may be conceding that general-purpose tools can do some tasks in legal.

It’s clear that TR is playing to its advantage in being able to offer strong assurances of correctness and trying to define the areas where that is necessary. It’s trying to convince its market the ability to defend the answers is critical.

The Pivot Does Make Some Sense

I can see why TR thinks this is so. As general-purpose models get better and can offer validity assurances that are sufficient for many practitioners, TR wants to grab that share of the market that believes it needs more sophisticated tools and protections. It’s a good play. It’s a market segment that believes it needs a product with the most bells and whistles

TR also may be gambling on the old notion that management won’t be faulted for buying and using products from a well-known industry leader: it used to be said that no one ever got fired for buying IBM. As long as that mentality prevails, there will be a good market for industry leaders like TR to offer tools that promise the sun, the moon and the stars. Plus, it gets at the desire of many firms to believe their work is the kind of work for which fiduciary AI is needed. Even if it isn’t.

But If I Were You

But despite everything TR is saying about the rebuilt CoCounsel, practitioners had still better, at least for the foreseeable future, verify outputs. And not become too reliant on what AI tells them, whatever the platform.

It’s noteworthy that TR is saying CoCounsel can now play the role of a senior associate. It was not that long ago we were told that AI could act like a beginning associate and be treated as such. Tomorrow it may be a junior partner. If we believe too strongly in AI, the logical extension will someday be a senior partner. Where does that leave us?

That may be where TR’s pivot is taking us. But if I were you, in the meantime, I would still check all AI cites.

Stephen Embry is a lawyer, speaker, blogger, and writer. He publishes TechLaw Crossroads, a blog devoted to the examination of the tension between technology, the law, and the practice of law.

The post Thomson Reuters Rebuilt CoCounsel: A Pivot appeared first on Above the Law.