On July 23, we gathered at Fabrik NYC for an evening on how AI is transforming the practice of law. The lineup was deliberate and intentional, designed to offer different perspectives: a product leader from consumer tech, a tenant lawyer and legal futurist who spent twenty years in housing court, a startup counsel advising founders, and a software engineer building compliance infrastructure.
They were not working from a shared script. They came from different ends of the profession and, in some cases, from outside it, which made it more striking that four of them arrived at roughly the same conclusion by four different routes.
“With these events, we are giving an audience to some of the most important innovators in this space, the lawyers and technologists who have thought deeply about how AI is transforming the nation’s legal practices. None of us can be 100% sure about what is going to happen next, so all of us need to keep an open-mind. The best way to adapt to these changes is to come to these gatherings and engage in conversations with the people who are actively inventing the future.”
– Lawrence Krubner, CTO, SCORE
The conclusion was not that AI replaces lawyers. It was that AI relocates where a lawyer’s value sits.
The bottleneck moved
Lizzy Palmer, our VP of Product, came to SCORE from product roles at Meta and Roblox, and started where product teams always start: who is the user, and what problem are we solving?
She expected the answer to be information, with AI arriving to summarize and organize.
Months of working alongside regulatory lawyers changed her read. The problem isn’t that lawyers need information. It is that finding information has gotten too easy, and deciding what deserves attention has gotten harder in exactly the same motion.
“The scarce resource isn’t information anymore. It’s attention.”
If the constraint is retrieval, you build a better search. If the constraint is attention, you build something that makes a judgment about relevance — and takes responsibility for what it leaves out. Those are different products.
Relevance is contextual, not absolute
But the challenge with making a judgment about relevance is that it varies by user and use case.
Ask a regulatory lawyer whether a given development is important and the answer starts the same way: it depends. On the client. On the jurisdictions they monitor. On the risks they care about. On what they happen to be working on that week.
A lawyer at a firm reads a new rule and asks: which of my clients does this affect? In-house counsel reads the same rule and asks: does this create exposure for my company? Same development, different question, different answer about whether it matters at all.
The development didn’t change. The context did.
Which means relevance can’t be a property of the development. It has to be a relationship between the development and a specific reader with a specific objective. Any system that assigns a universal importance score is answering a question nobody asked.
The value moves from the answer to the question
Kymberly A. Robinson, who founded Kymesq Law and advises founders from formation through exit, made a similar point.
“AI answers the questions you ask it. Lawyers are called counselors at law — we’re trained to identify the questions to ask.” Those questions depend on the client context.
Her examples were specific and compelling. What happens if this investor wants control? Have you protected the software before you launch? What does this term mean for your next round, or for your relationship with your cofounder? A founder facing a term sheet doesn’t know to ask most of those. That’s the whole job.
She also noted something that should interest anyone building tools for lawyers: her clients are unmoved by effort. They aren’t impressed that a contract took five hours to draft and three more to revise. They want to know whether the issue got found and resolved. One client’s line has stayed with her: I don’t need another person to tell me what the law says. I need someone who understands what I’m trying to build.
She was equally direct about the failure modes. Templates that don’t know your state, your entity, or your cap table. A skipped trademark clearance that turns into an expensive rebrand down the road. And confidential information pasted into a chatbot by people who haven’t thought carefully about privilege.
Cheaper doesn’t mean smaller. It means wider.
Two speakers approached the economics, from opposite directions.
Sateesh Nori spent twenty years as a tenant lawyer in New York City’s housing courts before becoming a legal futurist and senior research fellow at NYU. His numbers are the ones that should end any conversation about whether legal AI is a solution in search of a problem: 92% of the legal problems faced by low-income Americans go unaddressed each year. Counting the middle class, he puts the number of Americans with an unmet legal problem somewhere between 80 and 100 million. His book, Sheltered: Twenty Years in Housing Court, argues we will never train enough lawyers to close that gap.
Ethan Shea, founder and CTO of Regbase, gave the mechanism a name: Jevons paradox. When a resource gets cheaper, consumption tends to rise rather than fall, because uses that were never economical suddenly are. Coal in the 1800s; compute today.
The implication is that the interesting effects of cheaper legal work won’t show up in the work already being done. They’ll show up in work nobody attempted, because the price ruled it out.
Where the analogy breaks
Shea was also the most careful about limits, which is worth noting given that he builds this software for a living.
Software engineering transformed quickly because it has unusually strong feedback loops. You can specify precisely what correct means. The inputs are already digital. Verification is cheap, and you can run it a thousand times.
Legal work has none of that reliably. Correctness is frequently interpretive. Mistakes can be irreversible in a way a bad deployment isn’t. There are counterparties with their own interests. And the authority to say a thing carries weight independent of whether the thing is right.
He closed on a thought experiment rather than a prediction: imagine any legal matter could be resolved instantly, for free, and perfectly justly. What would people do with that?
What this means for regulatory practice
Regulatory work is where these threads pull tightest, because it is the area where the volume problem is already unmanageable and getting worse.
The teams we work with — at firms, in-house, and on the investment side — are not short of regulatory information. They have newsletters, trackers, alerts, and spreadsheets. What they lack is a defensible answer to a narrower question: of everything that moved this week, what actually touches my clients, my jurisdictions, my exposures?
That’s the problem SCORE is building against. It was insightful to hear four different perspectives converge on the same conclusion.
And Lizzy was specific about what building against this problem requires:
“These lawyers aren’t just our users, they’re our design partners. Every conversation has the possibility to change the product. If AI is going to change legal work, lawyers should be part of shaping that future.”
Book a demo →
About SCORE
Built by practitioners.
Powered by expertise.
Designed for decisions.
SCORE is a Strategic Compliance, Oversight & Risk Evaluator that leverages AI and technology, combined with human expertise and oversight, to transform global legal, regulatory and policy signals into forward-looking intelligence. Led by a former Baker McKenzie Principal Economist and a former Kirkland & Ellis Partner, SCORE serves corporations, law firms, and private equity funds.