There is a question circulating in every legal technology conversation right now, and it usually sounds something like this: if the foundational models keep getting more powerful, why would anyone pay for a vertical legal AI tool?

It is a reasonable question. But it is also the wrong one.

The better question, the one that actually determines how your legal team operates a year

from now, is not whether a general-purpose model can do legal work. It is whether your team

will get a curated, reliable, context-aware experience from it. Or whether they will spend their time building that experience themselves, one prompt at a time.

Ruli AI’s CEO Bryan Lee sat down recently with Anna Guo of Legal Benchmarks to discuss exactly this tension. What emerged was a clearer way of thinking about where the legal AI market is heading and what legal teams should be paying attention to as they make decisions about their technology stack.

The Restaurant Analogy That Reframes Everything

Brian offered an analogy during the conversation that captures the market better than most frameworks I have seen.

He described three tiers of the legal AI landscape as restaurants. At one end, you have the

foundational models, which provides you with a high-volume, quick-serve experience. Broad menu, fast delivery, but no one is tailoring the meal to your dietary needs or remembering what you ordered last time. At the other end, you have the premium, velvet-rope providers, which are impressive and exclusive, but built primarily for a different customer profile, one that has minimum commitments and a roadmap shaped by their largest clientele.

And then there is what Brian described as the boutique neighborhood restaurant. The place where the chef comes out, remembers your name, takes requests, and shapes the menu around what the community actually needs.

This is a vivid analogy. But what makes it useful (not just clever) is what it reveals about a structural reality in the market. The question for most in-house legal teams is not “which restaurant has the best kitchen?” It is “which restaurant will actually cook for me?" 

That distinction is everything.

The DIY Question Is Real. But the Answer Is More Nuanced Than It Appears.

Anna pressed Brian on a point that many legal teams are wrestling with right now: the rise of tools like Claude Code and Claude Cowork, where users can build their own skills, their own workflows, their own integrations without writing a line of code.

It is a fair challenge. If a GC can spin up a custom legal research workflow from a foundation model in an afternoon, what is the case for a dedicated platform?

Brian’s answer was grounded in something I see consistently across the legal teams I work with: most legal professionals do not want to become product builders. They want to practice law, just

faster, with better information, and with fewer manual steps. The early adopters, those comfortable assembling their own tooling (and with the time to do this), are a real segment. But they are not the majority of the market.

The majority wants curation. They want someone who has already thought through the information architecture, the source quality, the guardrails, the UX decisions that make the difference between a tool that technically works and one that your team actually uses on a Tuesday afternoon when a business unit needs an answer in two hours.

This is not a knock on the foundational models. They are extraordinary. But the gap between capability and usability, between what a model can do in theory and what a legal team will reliably get from it in practice, is where purpose-built platforms earn their value.

From Co-Pilot to Conductor: The Agentic Evolution

One of the most forward-looking parts of the conversation was Brian’s description of how Ruli AI’s product vision has evolved, and where it is heading.

Ruli AI did not start with a chatbot. The original vision was agentic: a digital paralegal that could triage, review, and orchestrate legal workflows autonomously. But, in 2024, the market was not ready. Back then, legal teams wanted a co-pilot, an assistant they could talk to, grounded in their institutional knowledge. So, Ruli took those requests and shaped its platform around what the community actually wanted, always with the platform architecture in mind.

Now, the market is catching up to the original vision. Legal teams that spent the past year building trust with AI-assisted workflows are beginning to ask the next question: what can we automate? Where can we move from human-in-the-loop to human-on-the-loop?

Brian described one example that illustrates this well: a public company using Ruli to automatically review compliance documentation against Fair Housing Act requirements, autoapproving low-risk submissions and escalating only the requests that require human judgment. This team had over 20,000 requests processed in a year.

That is not a chatbot. That is infrastructure.

And the trajectory Brian described, from the lawyer in the individual seat to the lawyer as conductor of an orchestra of agents, is one that resonates deeply with how I think about the future of legal work. The lawyer does not disappear. We just move to a higher-leverage position, directing systems that handle volume while preserving the judgment layer where our human expertise is irreplaceable.

What Legal Teams Should Take Away

If you are evaluating legal AI right now, there are three things from this conversation worth sitting with.

First, start with your use cases, not the vendor landscape. Brian’s advice mirrored what I hear from every legal team that has gone through a successful evaluation: isolate two or three things you actually need to solve, get buy-in from your team on those priorities, and test against them specifically. Trying to evaluate everything a platform can do is a recipe for decision paralysis (not to mention your time wasted).

Second, ask which vendor will prioritize your voice. This is the question that separates a good evaluation from a great one. If you are buying from a company whose primary customer is a 200-person law firm or a 50,000-seat enterprise, your feature requests are unlikely to shape the roadmap. If you are buying from a company whose core focus is in-house legal teams at your scale, you have a fundamentally different relationship with the product’s future.

Third, pay attention to where the product is going, not just where it is today. The shift from copilot to agentic workflows is not theoretical. It is happening now, in compressed timelines. The platform you choose today should have a credible path toward the orchestration layer your team will need in 18 months.

The legal AI market is maturing fast. The teams that navigate this moment well will not be the

teams who chose the most impressive demo. They will be the teams who chose the partner that understands how their team actually works and is building toward where legal work is actually going.