In this piece for SmartBrief's Practical AI series, Darrow's VP of Legal Intelligence Amit Weiss makes the case for a fundamentally different approach to legal practice — one that begins not with a client walking through the door, but with the data that signals harm is already occurring.
Weiss opens with a scenario designed to illustrate the core problem. Residents of a small Illinois town begin experiencing headaches, nausea, rashes, and tumors. Some post about their symptoms on social media, others file complaints in consumer protection databases, a few stories make local news. The signals are all there — but no one is connecting them. The source of harm, a nearby manufacturing plant emitting toxic waste, remains legally invisible. The affected residents have no recourse, not because the law doesn't protect them, but because the pattern that would make their case legible has never been surfaced.
This, Weiss argues, is the fundamental gap in today's legal system. Traditional legal processes rely on an individual reaching out for help — a model that is slow, reactive, and by definition excludes most people affected by systemic violations. Evidence for violations remains buried in disconnected data sources, unnoticed by the people harmed and undetected by legal teams. Without upstream tools, wrongdoing can remain legally invisible indefinitely.
Legal intelligence exists to close that gap. Weiss defines it as the structured process of using AI, data analytics, web intelligence, and legal expertise to uncover violations hidden in public data. At Darrow, she leads a multidisciplinary team of analysts, technologists, and lawyers who develop and deploy systems that process large volumes of publicly available information to surface patterns consistent with legal harm — across areas including environmental law, data privacy, antitrust, and product liability.
She describes the five-stage process her team follows: asset development, data collection, pattern analysis, legal validation, and operational planning. The model borrows from investigative disciplines used in national security and law enforcement, but produces a distinctly legal output — evidence structured to support a viable case. Three capabilities are essential to making it work. The first is cross-source data aggregation: signals are scattered across consumer reviews, environmental reports, financial filings, and complaint databases, and no single source tells the complete story. The second is pattern recognition grounded in legal context: not every anomaly is a violation, and distinguishing between what is merely interesting and what meets a legal threshold requires human expertise that most AI tools alone cannot provide. The third is structured, actionable output: a fully developed case assessment including factual background, jurisdictional strategy, class size estimates, and damages projections that allows an attorney to move quickly once a violation is confirmed.
Weiss is clear-eyed about the limits of AI. Large language models, clustering algorithms, and generative AI all play important roles in processing data at scale — but every detection at Darrow is reviewed and validated by its Legal Intelligence team before any case moves forward. The goal is machine-scale pattern recognition combined with human legal judgment, so the system can respond to risk with both speed and accuracy.
The piece closes with a call to the profession: justice should not depend on chance discovery. Legal intelligence offers a model for upstream practice — one that invites attorneys to think less like litigators waiting for clients and more like investigators asking where harm is already occurring.
