Unsurprisingly, there are a lot of lawyers at Harvey. In addition to conventional in-house counsel functions, we have former lawyers supporting every aspect of our company—from foundational AI research to client relationships and empowerment. My team, applied legal research, is one of the more unique applications of legal skills at the company. In this post, I explain what an applied legal researcher (“ALR”) is, why we have ALRs and how they help design, build, and improve the AI systems at Harvey.

When I joined Harvey, we had just gotten access to a large language model (“LLM”) that was miles ahead of anything publicly available. This LLM eventually became GPT-4. At the time, my mandate as head (and sole member) of applied research was to answer three deceptively simple questions:

These are all research questions, open-ended and unsolved. What makes them applied research problems is our approach to solving them. At Harvey, we don’t care about improving results on abstract benchmarks; we care about delivering tangible value to our clients. And we care about delivering this value today as much as—if not more than—we care about next year’s breakthrough (which usually results from the relentless pursuit of today’s problems, anyway).

Over the last two years, the scope of our applied research questions has expanded exponentially. We now ask, what can these models (OAI, Anthropic, Mistral, and more) do? The tools for making those models more performant now include optimizing prompt engineering, retrieval, fine-tuning, mid-training, agentic systems, and other facets of system design. Our client base has also grown well beyond lawyers and now includes many kinds of knowledge work professions, such as tax, finance, consulting, and private equity.

Despite this scope expansion, the applied research mandate has remained relatively constant: leverage all of these resources to make AI do more for professional services.

Delivering on the potential of AI to revolutionize professional services requires much more than legal expertise. Harvey now staffs product teams, designers, engineers, and machine learning researchers all committed to enabling lawyers to leverage the best AI possible. As we have grown this incredible and interdisciplinary team, we have also proven out the original thesis of Harvey: building the best AI for legal requires a close collaboration between the best legal and AI talent.

This is because teaching a model to do a complex legal task requires domain expertise at every step. Even answering the basic question of which legal tasks are worth solving with AI requires combining an understanding of lawyers’ pain points with an appreciation of AI’s capabilities to solve them. From there, legal researchers are essential for a number of functions, including the following:

As our team has grown, we have also found that lawyers are critical to our product team. ALRs now collaborate with product and design. They provide perspective on whether a product makes sense, whether it emphasizes the most essential information, and whether it delivers our industry-leading AI in a way that lawyers would expect and find accessible.

Today, Harvey employs nearly a dozen ALRs to support our AI and product functions. All of them share two key traits. The first is being elite lawyers from top law firms. The second is having no prior experience with AI. The second tends to surprise more than the first. However, ALR work at Harvey requires a strong, but unique set of AI fundamentals that we believe can be learned nowhere else but here. As such, the only expectation we have for incoming ALRs is a deep understanding of our clients’ challenges and the legal knowledge and judgment to solve those problems. From there, we can teach how to solve those problems using AI systems.

We asked our current ALR team about their experience, the skills they’ve built, and problems they’ve solved while at Harvey. Themes included the unique opportunity to build legal AI products in a fast-paced, exploratory, and intellectually rigorous environment. They described the role as a combination of some of the most enjoyable aspects of practicing law—problem-solving and devising creative solutions to novel challenges—with the immediate impact of product development. Concretely, product highlights include a diverse array of solutions including:

If these are problems you want to work on, the ALR team is growing. Not only are we always looking for more exceptional lawyers to join our team, but we are also hiring other knowledge work professionals to meet the expanding and diversifying demands for Harvey solutions. If you have worked on the hardest knowledge work problems and are interested in changing the way people solve them—join Harvey.