How to choose — and keep up with — an AI solution

This article was produced in partnership with LexisNexis

As AI technology continues to improve, it’s important to realize it’s not just the product that’s evolving — the skills of the lawyers using it must grow too. Taking a wait-and-see approach while awaiting new developments costs firms the opportunity to improve skillsets and workflows in the interim.

“The sooner the technology is in the hands of users and embedded in firm processes, culture, and business models, the more they will be able to maximize the value as the product improves,” says Samuel Puchala, President and General Manager of LexisNexis. “I'm not saying rush into a decision — do your due diligence. However, don’t sit on the sidelines and wait this out.”

How to choose the right AI tool

While there’s no one definition, at the highest level the right AI tool for the legal industry is one that provides the best output in the shortest amount of time, with answers grounded in authoritative legal databases of primary and secondary content. It should protect the privacy of the client and the sensitivity of the data, and its development should also align with ethical AI principles.

But the question then becomes, how do you choose one? Puchala’s biggest piece of advice is this: test the product, not yourself.

“There’s a whole skillset involved in prompting that’s new for most of us, and with legal Gen AI, you really have to know what you’re doing to maximize the tool,” he explains. “We emphasize this with customers, tell us what you want it to do, and let us help you develop the prompts and follow ups. In order to get to a good response, you have to go through that iterative process.”

Puchala likens it to dealing with an articling student or junior associate in that it’s not a partner-level resource capable of replacing you or your judgment but should be a resource capable of advancing your work product fast right out of the gate, while clearly showing potential to continue developing. The latter point is critical, as there’s a lot of “vaporware on the market right now — products high on vision and low on what it does today,” he notes. While the space is evolving quickly and the responses will continue to get better for years to come, “the product should be impressive today.”

“If you’re looking at something that has a fantastic roadmap but right now isn’t standing up to the test, that particular vendor is behind. It should be adding value right out of the gate.”

Puchala says privacy and security of data are also currently top of mind, and he repeatedly hears concern that the user’s search history and usage information is stored or used to train the model, and if safeguards exist to prevent exposure of client information through the generated content. 

“If you're kicking the tires on security and privacy these are paramount, and any other questions ultimately roll up into them,” Puchala says, adding that the other main consideration is one people don’t explore as much as they should.

People should get comfortable with how the models underlying the tool work to be able to know whether they can trust the architecture to, on a consistent basis, get them what they need. There are several quality LLMs that legal Gen AI tools sit on top of, and each one has strengths and weaknesses. But the high-level question is does the solution use a multi-model approach to optimize outputs, especially for different use cases?

Some models lend themselves more to answering a legal question, while others are more adept at assisting in the drafting of a document, for example, so a solution that combines outputs from various models is a huge asset. Its overall performance will always surpass that of any one model.

Lexis+ AI™ ticks all the boxes

LexisNexis is focused on supporting law firms through the decision-making process, from providing a guide on how to evaluate solutions to ensuring the product design of its own Gen AI technology ticks all those assessment boxes — and continuously exceeds expectations.

As a global company, the LexisNexis ethos is clear: improve decision making and outcomes and advance the rule of law around the world by making it predictable, reliable, and transparent. AI slots nicely into that mission and is nothing new at LexisNexis. Extractive AI has long been used in their existing products to surface insights and optimize results, and Gen AI is simply the next frontier.

Lexis+ AI is a single, integrated product grounded in industry-leading data, with industry-leading data security and attention to privacy. Hitting that sweet spot between comprehensive but simple to use, the idea is an all-in-one solution and, Puchala says, it’s simply “a continuum of the road we’ve been on for quite some time.”

Lexis+ AI was launched to the Canadian market on Thursday, July 25, 2024, following the successful launches of Lexis+ AI in the USA, Australia, the UK, and France. But true to LexisNexis’ nature, the goalpost is already moving.

Not only will the solution continue to evolve rapidly with improved response quality, upgraded features, and new use cases, but recently acquired Belgian legal tech company Henchman — a provider of enrichment data from document management systems for faster document drafting — will be integrated into Lexis+ AI soon. Its inclusion will enable LexisNexis customers to use their own reliable data and LexisNexis data as the basis for generative AI research and drafting - taking personalization to the next level. This cutting-edge technology seamlessly integrates with the tools legal teams use every day, providing a streamlined, superior drafting and negotiation experience.

“That’s a huge differentiator,” Puchala says. “Being able to use both our legal data bases and the firm’s internal database within the AI is a huge next step and a top priority for us.”

Gone are the days when, especially in legal research, a product was released followed by a few bug fixes, and then in a few years the new version would come out. It’s a very different cycle now, Puchala says, noting that Lexis+ AI in the U.S. was launched in October 2023, and it has already gone through iterations in the double digits.

“The product you see today is different from two weeks ago, and it will be different again in a month,” he says. “There’s a ton of value now even as it continues to grow — don’t miss out.”