Robert Ambrogi Blog
LexisNexis Legal & Professional today announced the opening of its Customer Innovation Lab in New York City, a physical space where legal professionals, LexisNexis engineers, and AI technology partners including OpenAI and Amazon Web Services work side by side to design, prototype and test new legal AI capabilities – sometimes iterating on an idea multiple times within a single hour.
The company says the lab – which actually opened about 1.5 months ago – offers a new model for how legal AI gets built and shipped. Rather than developing AI capabilities internally and asking customers to validate them after the fact, the lab brings customers into a working session, where LexisNexis engineers build against their specific pain points in real time.
Prototypes that clear the company’s standards for quality, accuracy and security can advance into development as capabilities within the LexisNexis agentic platform, either customized for the customer or scaled across Lexis+ with Protégé.
“For me, the biggest thing is we’re collapsing the distance between the people who know what the problem is and the people who can build the solution,” Greg Dickason, chief technology officer of LexisNexis Legal & Professional, told me in an interview for my LawNext podcast, which will air later this week.
“Putting them in the same room means that our technologists really understand the problems, which means that our ability to respond very quickly to where the market’s going is hyper-accelerated.”
A Different Approach to Customer Input
Over the years, I have sat through any number of company briefings – including by LexisNexis – in which executives told me that customer involvement was a core tenet of how the company develops products. So I asked Dickason to explain what is actually different with the lab.
His answer, in a nutshell, was cycle time. The traditional approach – which involves interviewing customers, going and building something, bringing it back for feedback, and then repeating the process – could take weeks or months per iteration, he said. In the lab, that loop collapses to minutes.
“Literally in the session in the lab, the customers are telling us their problem. We’ve already got the first solution, because we’ve listened and we’ve already pulled that together. We show that to the customer. They say to us, ‘Yeah, but it would work slightly better if this happened or that happened.’ And then we tweak it live in the lab,” Dickason said.
Just last week, he said, he saw a solution get changed four times in one hour as a result of customer feedback.
The idea is for everyone to be hands-on in real time, Dickason said. While one or two LexisNexis staff talk with the customer, engineers sit alongside, listening and coding, often using agentic coding tools such as Claude Code integrated into the LexisNexis product. At some point, they turn their laptops around and ask, “Is this what you meant?”
Dickason said the lab has been quietly operating for about a month and a half ahead of today’s launch, with the company using that time to refine its working cadence, including how a lab team of roughly 20 people hands off prototypes to an engineering organization of some 4,000.
For functionality-level features, he said, the path from lab prototype to live product “can be less than a month.”
Already Out of the Lab
The lab has already produced prototypes and designs that LexisNexis says are advancing into Lexis+ with Protégé capabilities, including:
- A customer-managed Protégé Vault for multimedia materials, allowing legal professionals to store and analyze documents, video, audio and images together. Using this tool, Dickason told me, a customer could upload a two-hour video deposition and ask when the witness showed surprise and what was being discussed at that moment, with findings traced to timestamps and source materials.
- Protégé Workrooms, secure shared spaces where teams can bring together documents, analyses, skills and AI-assisted work on a matter with outside counsel, clients, experts and other authorized participants, governed by permissions and access controls.
- LexisNexis-developed legal AI models, purpose-built for legal work using the company’s content and expertise.
The vault work, Dickason said, originated from a gap the company had heard about from customers, then took shape when customers came into the lab and described the actual queries they would run. He said that one of LexisNexis’s core technology providers, which he declined to name, was in the room with the customers, helping define what was technically possible.
Drafting As A Priority
LexisNexis says that the lab’s current top priority is legal drafting. The company cites research showing lawyers spend 40 to 60 percent of their time drafting and reviewing documents, more than on any other task.
Existing tools, Dickason said, get lawyers only 80 or 90 percent of the way there, leaving hours of work to reach a final product. The gaps he identified included conforming drafts to a firm’s own style, precedents and playbooks; flagging risk against a customer’s own standards; carrying context across related documents so a primary document and its supporting materials line up; verification of citations surfaced directly in the draft; and layering in litigation and judicial analytics.
Frontier Models, For Now
With Thomson Reuters recently announcing that its own proprietary model will begin replacing general-purpose models for certain tasks, I asked Dickason whether LexisNexis sees itself heading the same direction.
“Never say never,” he said, but for now the company’s view is that frontier models — he specifically mentioned Claude Opus 5 — combined with LexisNexis’s knowledge graph, retrieval infrastructure and prompting deliver state-of-the-art results.
The company has evaluated models continuously since its work with Amazon on Bedrock more than three years ago, he said, and already uses fine-tuned models for certain parts of the legal workflow.
“We’re going to make sure the highest quality is for us the number one bar,” he said.
Dickason also sees an agent-driven future in which AI agents increasingly act on behalf of lawyers — “disintermediating the user,” as he put it — and said LexisNexis is working on agent-to-agent and MCP interfaces to participate in that ecosystem.
Notably, he does not want third-party agents browsing raw content inside Lexis+ with Protégé, arguing that raw retrieval without the company’s grounding and tuning “generally doesn’t work.”
Who Gets In
Participation is open to “effectively anybody,” Dickason said, from CIOs and partners to junior associates, small firms, solos, and even judges, and whether or not they are current customers.
LexisNexis owns the IP in what comes out of the lab, which participants acknowledge before joining. Dickason said the lab is also set up for a separate track in which large firms building their own internal AI platforms can explore how LexisNexis might plug in.
Now that the lab has launched as an ongoing operation in New York, LexisNexis plans to expand it globally as well as to conduct on-site lab sessions at ILTACON and at customer locations. Firms and legal departments can learn more at www.lexisnexis.com/lab.
My full interview with Dickason will air later this week on LawNext.