Business Insider's Melia Russell visited Harvey cofounder Winston Weinberg for an exclusive look at the legal AI startup, finding that Harvey is training specialized models to handle dense legal text, automate routine legal work, and compete in a suddenly crowded market. Law firms that adopt these tools could see shifts in how associates spend their time and how quickly clients receive analysis.
Why law is a natural fit for generative AI
The legal profession runs on text - statutes, rulings, contracts, and decades of case law. Large language models thrive in this environment. Harvey, cofounded by Weinberg and Gabe Pereyra, uses that mountain of legal data to build AI assistants that can parse and summarize complex documents. Associates who once spent years learning to interpret these texts now see parts of that work handled by software.
At a recent tech event, Weinberg's team included a practicing lawyer who, instead of advising clients, trains Harvey's models to produce more accurate legal outputs. The role signals a shift: legal expertise is moving from the firm library into AI training pipelines. The high-quality, structured data that law generates makes it an obvious training ground for generative models.
How Harvey is tackling new competition
Harvey is not alone. The legal tech market has heated up, with multiple startups building AI tools for document review, contract analysis, and litigation support. During Russell's visit, Weinberg shared details about the company's latest moves to stay ahead. Those include refining its models with more precise training data and expanding the types of legal tasks its software can handle beyond simple summarization.
Startups like Harvey are creating specialized tools under the broader category of AI for Legal, automating busy work that previously bogged down junior lawyers. The push mirrors similar efforts in consulting, where firms like Deloitte and PwC are investing in AI agents to reduce thousands of hours of manual slide creation and analysis.
Why this matters for Legal Professionals
AI tools that read and interpret legal text won't eliminate the need for attorneys, but they will change which skills get rewarded. Understanding how these models are trained, what they can do reliably, and where they fail becomes part of a lawyer's toolkit. The lawyer at Harvey isn't just debugging code - she's helping define what "good" legal reasoning looks like for a machine. Legal professionals who grasp that process early will be better positioned as their firms adopt similar tools.
Your membership also unlocks:
