Here’s what I’ve learned from building legal AI: the technology should work around the way lawyers practice, not the other way around. That means connecting powerful AI models with authoritative content, specialized tools, and the context of the matter. The result is AI that lawyers can rely on for better-informed decisions, stronger work product, and better outcomes for their clients.
Frontier models – the most advanced general-purpose AI models – have improved remarkably fast, and nearly everyone can now access them. But a model is only the engine. On its own, it doesn’t know the matter, the authorities, or what good legal work looks like. The right system around it supplies all three.
At LexisNexis, we call that system the legal harness. It’s agentic, meaning it can take steps toward a goal, not just answer a question. It brings everything the work requires together around what the lawyer is trying to accomplish.
Legal work calls on many capabilities
Consider a lawyer who asks: “Help me evaluate the strongest arguments for this motion.” Answering well means far more than generating text. The system needs to understand the facts and documents in the matter, research the relevant law, identify controlling authority, look at how the court and the judge have handled similar issues, weigh contrary authority, draft the argument, and confirm that every cited authority exists and is still good law.
Each of those steps may call for a different model, source, or tool. Today, many lawyers stitch those pieces together by hand, moving between research products, databases, AI tools, and document environments, and carrying context from one to the next. That’s time spent managing technology instead of practicing law.
The harness is built to handle that coordination for them.
How the harness works around the lawyer
Under the hood, the harness runs a simple loop: perceive, decide, act, observe. It assesses the request and the available context, decides which capability fits, takes an action, checks the result, and determines what comes next.
That’s the key shift. Rather than pushing every request through a fixed sequence of steps set in advance, the harness adapts as the work unfolds. Depending on the task, it might select an AI model, retrieve authoritative LexisNexis content, apply a Skill (a ready-made AI capability for a specific legal task), run analytics, work with the customer’s own documents, or call on an AI agent built for a specific job.
The lawyer sets the objective. The system works out how to get there. And the lawyer stays in control of the legal judgment throughout.
The right model for each task
No single model is best at everything, and the leading models change constantly. One may excel at certain kinds of reasoning or drafting; another may be better at extracting key details from documents. That’s why we take a multi-model approach rather than building around a single provider. The harness lets us evaluate each advance and apply it where it adds the most value, without asking legal professionals to make those choices themselves.
In our own testing, models that improved on general AI benchmarks didn’t always improve as much on real legal work. Much of the difference came from the system around the model: how it connects to information, tools, and other capabilities.
A better model helps. The right models inside the right system are what change the work.
Where 200 billion legal and business documents and records come in
Content depth is what makes the difference. The harness operates within the LexisNexis Intelligence Engine, which powers Lexis+® with Protégé™. It can draw on more than 150 years of attorney-curated content and more than 200 billion LexisNexis legal and business documents and records; Shepard’s® Citations and Shepard’s® Verify; the firm’s own knowledge; insights and analytics; and a growing library of Skills, tools, and AI models.
As leading models become widely available, the system around them matters even more. At LexisNexis, this means powerful models connected to everything a lawyer relies on, all coordinated around their work. That combination is what turns a plausible answer into work product a lawyer can verify, cite, and defend.
Trust that keeps pace with capability
The legal harness also provides a common foundation as AI keeps changing. New models can be evaluated and brought in. New Skills and tools can be added. Improvements to the architecture benefit every product and practice area at once, so nothing has to be rebuilt feature by feature. For lawyers, this means new capabilities show up in Lexis+ with Protégé.
For lawyers, this means the latest AI advances arrive directly in Lexis+ with Protégé, ready to use in the work they’re already doing and held to the same trusted standards.
Every capability added to the harness is measured against the same standards: safety, accuracy, completeness, authority, and whether results are grounded in reliable sources. That consistency means trust never trails capability.
Built around the lawyer
Models will keep getting more powerful, and lawyers will benefit from every advance. The greatest value comes when those models work within a legal intelligence system built around the lawyer.
That system brings together authoritative information, specialized legal capabilities, and the context of the matter as the work unfolds. It grounds and verifies the results while keeping the lawyer in control of the legal judgment.
That’s how powerful models become better legal work: not by asking lawyers to adapt to the technology, but by building legal AI that works around them.
To learn more about the LexisNexis Intelligence Engine, read The Legal AI Journey: From Task-Specific Tools to Integrated Intelligence. To explore Lexis+ with Protégé, visit www.lexisnexis.com/ai.
# # #
For media inquiries about this post, please contact Jill Van Nostran,
jill.vannostran@lexisnexis.com