LexisNexis Legal & Professional today (24 August) unveiled new capabilities for Lexis+ with Protégé, introducing an ‘agentic harness’ that enables legal professionals to leverage multiple tools in one workflow within a Legal Intelligence Engine that now retains context.
Speaking to Legal IT Insider, Greg Dickason, executive vice president and chief technology officer at LexisNexis, said: “We’re packaging up all the ways in which we’re working into the LexisNexis Legal Intelligence Engine, but the piece that I think is probably net new is that context, the ability to carry context forward.”
Advertisement
The new functionality means that research and drafting no longer have to happen as separate steps: Dickason said: “You can say to the agent, find out what the latest law is about this and draft me a document for my client. And it’ll do both at the same time.”
Alexandra Smyth, executive vice president and general counsel for the LexisNexis Legal added: “I think the benefit of this is the research is embedded within drafting. So you don’t have to separately do research and then turn around and draft it using the research. You can just do it all in one spot. So if you’re drafting something that requires research, it’s now no longer two steps.
“In the past, if I wanted to ask a question about the SEC, I would go into Intelligize and look it up there. If I wanted to ask a question about what a judge had done in the past, I look in Lex Machina. Whereas now this harness brings it all together. So I can just ask it a question and it knows which one to call on. It gives me the answer and then it allows me to put that answer into Word, PowerPoint, Excel, whatever, without having to do all of that separately.”
Smyth points out that you can ground the new engine in your policies, precedents and risk appetite, commenting: “We’ve got the LexisNexis corpus, which is fantastic, but we’ve also got what the RELX policies are or what we’ve done previously. So you can tailor it to your own bespoke way of using it.
“What I really care about is getting a piece of output that I can actually trust and not worrying about whether it’s been hallucinated. We’ve seen that loads of times, I think, not with our products, but with lawyers in general, being burnt by legal AI. And so what I like about our product is you can easily check everything it gives you. There’s links to the citations.”
Dickason was clear that context is walled off by matter, so work done for one client cannot leak into another. He described the system as “a fully agentic harness” that checks its own output against LexisNexis’s Shepard’s citation database rather than relying on the underlying model alone: “It doesn’t matter how good your agent is, it can always make up something. You’re never going to get to 100% non-hallucination. So you’ve got to put the harness around it.”