Today, Thomson Reuters joined the race to announce its entry into the Large Language Model (LLM) marketplace and what it generally plans to develop in the future. The Company also shared its vision for the future through generative artificial intelligence. Finally, the Company announced a partnership and new plugin with Microsoft 365 Copilot, Microsoft’s AI offering. Thomson Reuters states, “This integration will bolster efforts for redefined professional work starting with legal research, drafting, and client collaboration.”
The Thomson Reuters announcement comes on the heels of LexisNexis’ recent announcement of its plans in the space. Other legal vendors like Casetext, LiquidText, and others have also jumped in recently.
These announcements—especially by Thomson Reuters and LexisNexis–are a bit unique in the legal marketplace. They are unique in what they are or are not now offering. And they are unique in how they see the implementation of generative AI tools.
But the uniqueness of the LLM tools makes these announcements sound different.
But the uniqueness of the LLM tools makes these announcements sound different. The whole LLM concept seemed to burse on the scent out of nowhere a scant five or so months ago. (Yes, the models existed before then, but the ability of the models to change how we do things was not well known until after November 2022). And the development of the models since then till now has been at 120 mph. Rapid development, exponential improvement, and change are the hallmarks of the models. And we don’t yet know precisely the long-term and short-term impact.
Secondly, how we humans relate to the models is fundamentally
different. Getting to the answer is not a search but a dialogue. It is conversational, much like asking a partner how to approach a legal knot you are trying to unravel. Both these characteristics—rapid and uncharted development and the new way we relate to the model–change how the vendors’ plan, what they say, and how they view implementation.
Integration. Integration. Integration.
In response to a question at the tail end of the Thomson Reuters briefing, David Wong, Thomson Reuter’s Chief Product Officer, emphasized that the goal for Thomson Reuter’s LLM is to bring together the knowledge and capability of the various Thomson Reuters products to answer inquiries. And the lines between the different Thomson Reuters products—Westlaw Edge, Precision, Practical Law Dynamic, Legal Doc Review, Legal Drafting, etc.—will eventually need to be blurred to maximize the abilities of generative AI. The goal, said Wong, is to create a consistent experience.
When I thought about it, I could see the logic behind this. When you ask your partner to help you with a legal problem, you don’t expect them to talk about how they would solve it from one batch of knowledge and ignore other knowledge or experiences they have. Instead, you expect the partner to offer insight from all the sources they may be familiar with. The power and promise of LLMs are that they can respond as your partner should.
There is a difference between LLM responses and pure search responses. It’s the ability of LLMs to understand natural language and then pool knowledge across fields that make them so powerful. So, product lines and siloing will become a real disadvantage as the models evolve.
Did I Mention Integration??
The LexisNexis and Thomson Reuters announcements were similar in another respect. Both are relying on Microsoft to help power their offerings. And that’s not surprising, given how ubiquitous Microsoft is in the legal community. It also reflects that Microsoft sees itself as the key player in legal. Its model is not to offer standalone products. Instead, it plans on being the integrating force that powers what other legal players are offering.
There’s a reason that neither LexisNexis nor Thomson Reuters cozied up with Google, which offers products that could compete with what Microsoft offers. In legal, Microsoft is king and will likely direct where and how the legal tech market goes in the future.
But They Really Didn’t Tell Us Much
A fair amount of criticism has been directed at LexisNexis and Thomson Reuters for offering something, the details of which remain a little sketchy. Details they both say will be developed later. The standard read is that both companies just wanted to jump on the bandwagon. That they just wanted to say something—anything—about LLM offerings, even if they didn’t have much to say. And yes, there may be some truth to that.
The generative AI field is developing at warp speed, and its impact and direction are still murky. So offering something too soon and too detailed may be a big mistake.
The generative AI field is developing at warp speed, and its impact and direction are still murky. So offering something too soon and too detailed may be a big mistake. The players may be right to proceed cautiously while they assess where LLMs are going, what they can do, and what customers want and expect them to do.
Even Google, which also recently announced a big generative AI initiative, was short on details. What’s funny is that when Google or Apple announces something new and is unclear on the future details, people aren’t surprised and even applaud.
Legal tech vendors, though, have traditionally had to take into account that lawyers expect products to do everything promised. Perfectly and immediately. If the product doesn’t do that, it’s quickly relegated to the trash can. So many of us who comment on the legal tech field may be surprised and a little skeptical about the kind of announcement that leaves details yet to be determined.
Generative AI is unlike anything we have seen before. It will change the legal field in ways we haven’t ever seen
But generative AI is unlike anything we have seen before. It will change the legal field in ways we haven’t ever seen. What it can do and what it will impact are yet to be fully or even partially determined. So promises about the future may be problematic.
Integration will be the name of the game as we shift to a conversational natural language tool. Harnessing that integration across product lines and vendors can’t yet be fully known and will take some creativity.
Perhaps we are seeing how legal tech vendors will approach these new realities generative AI brings.