To the folks at LexisNexis, the potential for generative AI to change how legal research is conducted quickly became obvious following the launch of OpenAI’s ChatGPT platform.
“The industry sort of had a moment where we saw the impact on business and society, frankly, with what ChatGPT could mean,” says Simon Weierman, Senior Director, Segment Management. “It became clear we could leverage this technology to deliver the insights that we offer in a more effective way to our clients.”
LexisNexis’ latest generative AI tool, Lexis+ AI, is now generally available, with firms such as Baker McKenzie, Reed Smith and Foley & Lardner helping LexisNexis work out the final tweaks through an extensive Commercial Preview program.
Above the Law recently spoke with Weierman about how Lexis+ AI will reshape legal research, making the process quicker and insights more accessible.
A New Generation Of Legal Research Tool
Machine learning and artificial intelligence are nothing new for LexisNexis, Weierman says. Since their beginning, the company has served legal practitioners with access to massive amounts of information, with users conducting hundreds of thousands of searches a day.
LexisNexis patented and offered natural language machine processing for legal search in the 1990s. “We’ve been using large language models to improve our products since at least 2018,” Weierman adds.
What changed with ChatGPT and similar AI products? According to Weierman, one of the biggest shifts involved the role of the researcher.
“ChatGPT provides an interface where the user becomes the programmer,” he explains. “It’s not this thing in the background that you can’t can’t touch and you don’t really know how it impacts you. Now, you can actually directly work with the AI as an end user.”
“Lexis+ AI provides a conversational interface that offers a way that is much more natural to get the answers and insights out of our platform.”
In a sense, Lexis+ AI is the natural successor to the natural language search that remains a core feature of Lexis+. It’s the newest way to rise to one of the oldest challenges in information systems: delivering relevant, accurate information to users as efficiently as possible.
A Faster, More Nimble Research Process
For years, the first step in conducting legal research involved casting a wide net and combing through those initial results, and continuously filtering and adding elements to find what is most relevant to the matter at hand. To get the best results, users have needed to master search techniques & strategies. Users also had to know in which databases their answers would be found.
Lexis+ AI, however, allows users to start with their specific context — just like you would if asking a question of a knowledgeable colleague, which also plays to the best practices for Large Language Model prompting.
According to Weierman, the conversational approach allows users to be nimble and more easily navigate roadblocks encountered during the research process, as well as worry less about which resource they should search to find the best answer.
With Lexis+ AI, “I can prompt a system in a way where I’m going back and forth more naturally and more rapidly than with the traditional search experience,” Weierman says. “When you’re starting broadly and then narrowing things down, it’s an extensive process… and if you’re not finding what you need, you can’t pivot nearly as easily as you can through this interaction.”
Weierman adds that Lexis+ AI responds particularly well to queries that include situational context, and that the team behind Lexis+ AI designed the tool to create outputs that are also sensitive to point of view an attorney is representing, the goals behind their research matter, and the ideal format for a prompt’s output.
Confidently Research, Draft, And Summarize
At the core of Lexis+ AI’s function are, of course, the tasks necessary for the day-to-day practice of the law.
Currently, this includes drafting documents (including letters, legal memos, email, arguments, & clauses), a conversational search function, document summarization (including case law opinions), and analysis & summarization of documents uploaded by the user.
These functions draw upon the vast libraries managed by LexisNexis — primarily case law, statutes, Practical Guidance, Matthew Bender treatises & practice guides, court rulings and agency decisions to start — with new content and data becoming available as the product continues to be developed.
And, if the user’s research scenario and search doesn’t fit with Lexis+ AI, the rest of the LexisNexis product suite is there to step in to aid the user.
“You still have access to the traditional, and still very powerful, search resources: Legal analytics, our Practical Guidance, Legal News and more,” Weierman says.
Importantly Lexis+ AI reveals information to the user about the sources used to generate output, making the process of verifying AI-supported research transparent and more efficient.
The Right LLM For Your Research Needs
One of the most interesting and unique features of Lexis+ AI is its seamless integration of several large language models (LLMs), with the program selecting the best model based on a given prompt.
“We take this multi-model approach because, in our testing and research, what we found is different models performed differently depending on the use case,” Weierman explains.
“Depending on the prompt, we could be using GPT-4, or using Anthropic’s Claude 2, and as the user continues their research, they are likely to leverage features using extractive AI powered by other LLMs such as Google Bard to connect our data and improve the relevance of our results & recommendations.”
Each of these models has been further fine-tuned for legal research by the Lexis+ AI team. For users, this means Lexis+ AI will use the most effective model for a particular prompt and return the best results available — something that a product that relies on a single LLM might not accomplish as effectively.
To learn more about Lexis+ AI and sign up for a demo, visit www.lexisnexis.com/ai. For more on the ethics of AI-based legal tools, take a look at the Responsible Artificial Intelligence Principles developed by LexisNexis’ parent company, RELX.