LexisNexis CTO Greg Dickason Applies AI Pragmatically

LexisNexis CTO Greg Dickason says the company’s AI tools are still heavily dependent on human oversight to avoid costly errors in courtroom settings.

Human checks remain central to AI deployment

In a recent interview, Dickason explained that LexisNexis continues to involve practicing attorneys in testing its AI models. The firm’s legal‑research platform processes massive data sets, but when the system misidentifies a case or fabricates a citation, judges have taken notice and called attorneys to account.

“Quite a few of them have upskilled into the data science stream,” he noted, referring to the roughly 4,000 engineers who develop the firm’s AI internally. By keeping development in‑house, they aim to maintain the context and precision required for legal work.

The validation process pairs AI models against each other and then asks a “couple of hundred lawyers” to evaluate the outputs. This double‑check helps the system “rethink its answers and potentially realize context it overlooked,” according to the report.

AI’s growing role in the LexisNexis ecosystem

AI currently powers about 30 % of LexisNexis’s services, and Dickason expects that share to rise to 70 % over time. The technology speeds up research that once required junior lawyers a full week; now, the same work can be done in minutes.

He compared the firm’s approach to “an open‑book exam with tens of thousands of legal documents” that are distilled into one or two pages for the model. By contrast, generic large models that ingest a million documents often become “lost because they cannot digest that much content and use it effectively.”

Beyond research, engineers are “doing much more coding and building using agents, like Claude Code, Codex and those kinds of platforms.” This shift reflects broader changes in how the technical staff create and maintain its products.

Despite the speed gains, Dickason cautioned that AI “clearly helps them do better — but not perfectly.” Errors can be subtle, especially when a model gradually introduces misinformation that then becomes entrenched.

He likened the technology to “a brilliant mechanic that can absolutely fix any car whatsoever but quite often forgets where the garage is and doesn’t know where the car keys are.” The analogy highlights the need for continual human verification.

In practice, the workflow includes a “double‑checking process” that pushes AI to get “unstuck” and refine its answers. This step is essential to prevent hallucinated cases from slipping into legal filings.

LexisNexis had already integrated AI into its search capabilities before the public hype surrounding ChatGPT. After broader market attention, the firm rolled out a customer‑facing AI model that supports both research and drafting directly within its interface.

Human oversight remains essential.

Looking ahead, the balance between automation and human expertise will likely shape how legal professionals adopt AI. While the technology can reduce the time needed for case analysis, the industry must remain vigilant against the risk of unchecked errors that could undermine courtroom credibility.