Despite the fact that ILTACON ended greater than per week in the past now, I’m nonetheless digging out from the deluge of briefings and conferences I attended there. Certainly one of specific notice was a media briefing with LexisNexis executives through which they offered additional data on the corporate’s AI technique.
Simply forward of ILTACON, LexisNexis had introduced the launch of Protégé Basic AI, as I reported on the time. This launch expanded its Protégé synthetic intelligence platform to incorporate safe entry to a number of general-purpose AI fashions alongside its present legal-specific AI instruments.
On the ILTACON media briefing, LexisNexis executives detailed new options of its Protégé AI platform, offered extra particulars on the rollout of the brand new Basic AI software, and mentioned the corporate’s broader imaginative and prescient of delivering what it calls “courtroom-grade AI” to the authorized occupation.
Sean Fitzpatrick, CEO of LexisNexis North America, UK and Eire, opened the session by describing the second as unprecedented by way of each technological change and monetary funding.
Associated: On LawNext: How LexisNexis and Harvey Are Partnering to Reshape Authorized AI, with LexisNexis CEO Sean Fitzpatrick.
“The quantity of funding that’s going into these foundational fashions is a whole lot of billions of {dollars}, and we’ve by no means seen something like this,” Fitzpatrick mentioned. “That is the largest expertise spending cycle within the historical past of enterprise.”
With prices for big language fashions dropping by greater than 99% since 2022, he added, the economics now make it doable to function these techniques at scale.
To that finish, Fitzpatrick described his firm’s imaginative and prescient of creating AI assistants extensively accessible.
“Our imaginative and prescient is that each lawyer goes to have their very own digital AI assistant, and it’s going to be personalised to them,” he mentioned. “It’s going to grasp their follow space, it’s going to grasp their jurisdiction, it’s going to grasp their fashion, their preferences, it’s going to have entry to our authoritative authorized supplies, and it’s going to have entry to their very own inner work product.”
Constructing ‘Courtroom-Grade AI’
Fitzpatrick mentioned that whereas consumer-grade AI instruments are extensively used amongst authorized professionals, they lack the privateness, quotation rigor, and reliability essential for authorized follow.
LexisNexis, he emphasised, is dedicated to creating “courtroom-grade AI,” which it defines by six ideas:
- Grounding in authoritative authorized content material. “LexisNexis has 160 billion paperwork and data in our system,” Fitzpatrick mentioned, “and in order that’s what we use as grounding information once we do our AI.”
- Verifiable supply citations. Shepard’s offers the flexibility to confirm that citations are correct and updated, he mentioned.
- Steady updates with no data cutoffs. Fitzpatrick famous that GPT-4o had a data cutoff of 2023. “We add 2.5 million paperwork and data to the system each single day from 50,000 totally different sources.”
- This has two facets, Fitzpatrick mentioned: One is transparency again to the authority, so you’ll be able to confirm for your self. The second is understanding what’s occurring contained in the black field – what’s the reasoning and logic.
- Bias mitigation. Gold-standard safety and privateness.
- Safety and privateness. It needs to be disconnected from the foundational fashions so the info isn’t getting used for studying or coaching.
“That is about constructing AI that attorneys can belief in a courtroom setting,” Fitzpatrick mentioned.
Protégé on the Heart
Central to that effort is Protégé, the corporate’s AI assistant. By 2028, he mentioned, Protégé will automate 15-20% of what attorneys do at present.
However he added that LexisNexis isn’t ready till 2028 to ship on that imaginative and prescient. Protégé already automates most of the duties attorneys routinely carry out, and it’ll proceed so as to add extra capabilities.
“Consider all the various things that all the several types of attorneys do, there are tens of 1000’s of duties that must be automated,” he mentioned. “We’re going to proceed to automate totally different duties alongside these workflows, so Protégé will get an increasing number of highly effective over time.”
Protégé is increasing on a number of fronts, Fitzpatrick mentioned:
- Geographically: It’s already stay within the U.S., U.Ok., Canada, Australia, Hong Kong and New Zealand, with extra rollouts coming to South Africa, France, and Austria.
- Throughout merchandise: It has been built-in into Lexis+, Lex Machina, Intelligize, and different instruments.
- Throughout the ecosystem: LexisNexis has added integrations with Microsoft Phrase, Outlook, Groups, and Copilot, in addition to a brand new reference to Harvey that permits Harvey customers to question Lexis AI immediately.
Agentic Workflows Demonstrated
Through the media briefing, Serena Wellen, vp of product administration, demonstrated Protégé’s agentic workflows, which she mentioned have been designed to extend transparency into the method and provides attorneys extra management over analysis and drafting.
In a single instance, Wellen confirmed how Protégé can break down a fancy question involving an airline’s responsibility to an unaccompanied minor into sub-prompts that the consumer can overview and edit earlier than the system delivers outcomes.
“Legal professionals can overview each quotation, take away irrelevant ones, and make sure the remaining response displays their judgment,” Wellen mentioned.
She additionally gave a preview of a judicial workflow at the moment in improvement, which permits judges to add case information and obtain draft bench memos or opinions. Judges who’ve examined it, she mentioned, in contrast the drafts to the work of “a extremely good legislation clerk” and praised the time financial savings of decreasing weeks of labor to minutes.
“A few of them even advised us that some elements of the drafts had been extra thorough than what they normally noticed.”
Protégé for Basic AI
As famous above, a significant latest announcement from LexisNexis was Protégé Basic AI, a safe workspace that lets customers entry fashionable shopper AI fashions equivalent to GPT-5, Claude Sonnet 4, and OpenAI’s O3 reasoning fashions from inside the LexisNexis setting.
David Ganote, vp of product planning, mentioned the transfer responds to the truth that two-thirds of attorneys already use shopper AI instruments equivalent to ChatGPT or Claude, typically for analysis, brainstorming or drafting.
“However they’re placing privileged information into open fashions that aren’t designed to guard it,” Ganote mentioned.
Protégé Basic AI, against this, presents enterprise-grade safety and incorporates LexisNexis’s Shepard’s quotation agent, which flags unverified or hallucinated citations. It additionally permits customers to toggle clearly between Authorized AI, grounded in Lexis content material, and basic AI, higher suited to exploratory or artistic duties.
“There are many issues that basic AI does very well that don’t require authorized authorities,” Ganote mentioned, pointing to examples equivalent to explaining ideas to a jury or drafting advertising and marketing content material.
Selective Method to Fashions
Jeff Reihl, govt vp and chief expertise officer, mentioned that the corporate’s improvement of brokers isn’t just a PR ploy.
“We don’t construct brokers simply so that you simply guys can write about us constructing brokers,” he mentioned. “Brokers are terribly highly effective expertise and we’ll use these to serve a function for our clients.”
Reihl additionally defined the technique behind the corporate’s alternative of AI fashions, saying that it maintains a multimodal technique, selecting the very best generative AI mannequin for every use case relatively than chasing the latest launch.
Though the corporate had early entry to GPT-5, he mentioned, it had not but been deployed in Lexis merchandise (at the least as of the press briefing).
“As of proper now, the efficiency isn’t higher, so we received’t use it,” Reihl mentioned. “We don’t put a brand new mannequin into our product simply because it’s the newest and biggest.”
LexisNexis has direct partnerships with Anthropic, OpenAI, AWS, Microsoft, and Mistral, giving it early entry to fashions and enter into their improvement roadmaps.
Aggressive Panorama
Because the briefing concluded, Fitzpatrick was requested concerning the broader aggressive panorama, and whether or not Microsoft or Google may finally come to dominate authorized AI. He argued that LexisNexis’ mixture of authoritative content material, safe AI, and workflow integration offers it an edge.
“It’s plenty of work to attempt to pull collectively the gathering of belongings that we’ve got,” he mentioned. “You may give anyone a billion {dollars} they usually wouldn’t be capable to do it.”
He additionally gave an replace on the partnership between LexisNexis and Harvey, which partially concerned the co-development of AI workflows for attorneys. I requested him if there had been replace on that since I spoke to him in July for my podcast.
“It’s developed since we talked about it,” he replied. “It’s actually coming collectively.”
He mentioned clients have been testing it and they’re constructing on the suggestions from these clients. “Actually, actually wonderful suggestions on it, so we’re very excited.”
Fitzpatrick closed the briefing expressing confidence in LexisNexis’s persevering with management in creating AI for attorneys.
“There’s plenty of room for individuals to innovate round these items, however I do suppose that main gamers like LexisNexis which have the content material … do have a bonus.
“From a expertise standpoint, it’s actually exhausting to keep up a aggressive benefit if it’s simply expertise. And in order that’s why we attempt to take our content material, embed it in our expertise options, after which embed these expertise options into the workflows of our clients.”
