American legaltech Harvey wants to transform the legal practices of law firms in the era of generative AI. Its president and co-founder details his ambitions to JDN.
JDN. You co-founded the Harvey platform in 2022, after having been an AI researcher at Meta and DeepMind. What convinced you of the opportunity to apply generative AI to the legal sector ?
Gabe Pereira. I would say there were two key moments. The first took place by testing GPT-3 on legal issues found on Reddit, including disputes between landlords and tenants. We had lawyers review the responses to evaluate them, and we realized how well these models already performed. Winston Weinberg, my co-founder and former roommate, was an associate at O’Melveny & Myers at the time and was thinking about how to use AI for legal research and writing tasks. Together, we had the idea of creating the Harvey platform in 2022. Then, there was a second click when we obtained early access to GPT-4. We were struck by his abilities, particularly in the area of programming.
Concretely, what are the main functionalities of Harvey and in which legal areas is your platform most useful? ?
I view Harvey not as a set of tools for individual tasks for lawyers, but rather as an infrastructure for firms as a whole. These large firms manage all kinds of very specialized client files, whether litigation, mergers and acquisitions, fund creation, etc. Each requires complex tools. What we’re building is a platform where an entire team of lawyers working on one of these cases can use generative AI combined with different tools.
Concretely, Harvey offers four main functionalities. First, Assistant, which allows you to ask questions, analyze documents and write faster thanks to specialized AI. Then, Vault, a collaborative space allowing you to store, organize and analyze legal documents in bulk in a secure manner. We also offer Knowledge, to perform legal, regulatory and tax research with precise citations. Finally, Workflows allows you to run predefined workflows or create custom ones, adapted to the specific needs of each practice.
Faced with those who qualify certain AI solutions as simple “ChatGPT wrappers”, is your differentiation ultimately more due to the collaborative infrastructure than to the models themselves? ?
A significant part of the platform is in fact aimed at IT departments and administrative teams of firms, with features dedicated to governance, security and data confidentiality. Harvey also covers collaboration with end customers. The platform allows you to share information and documents securely in order to work more efficiently on common files.
Simply put, what we’re building is all the infrastructure that law firms and large corporations need to use these AI models at scale on sensitive legal cases. This includes security, data privacy, governance, access management, as well as the orchestration of different tools and processes. If we take an example like Salesforce, the value of the software does not come from the fact that it stores customer data, but from the ecosystem and workflows built around it.
Can you give concrete examples of using the platform and explain how firms measure return on investment?
The ROI differs depending on the different areas of legal practice and for each specialty. In litigation, for example for cases carried out for Fortune 500 companies, the main issue is often the speed of execution. The ability to quickly analyze large volumes of documents, structure legal reasoning or draft pleadings more efficiently. In other areas, such as fund creation or private equity, the indicators are different. Generally speaking, we want to help firms improve the profitability of their activity.
What is your business model today and what are the major technological building blocks on which the platform is based?
Today, our model is based on subscription pricing, generally billed based on the number of users. We have built AI systems that use a mix of foundation models, combined with search orchestration systems, RAG and other techniques. The platform allows you to work both from a web application and directly in everyday tools like Microsoft Word. We work with several foundation model providers and different cloud infrastructures to adapt to customer needs.
Your “Workflow Builder” feature allows you to create automations for legal tasks. How does it actually work? ?
The objective here is to be able to use generative AI in a structured way to accomplish certain recurring tasks. We could compare this functionality to those of a tool like Zappier, with the difference that it is specialized for legal use cases. A workflow could make it possible, for example, to automatically generate a due diligence memo from a data room. The goal is to help firms transform high-volume processes into standardized workflows capable of operating at scale.
Concretely, a lawyer can describe in natural language what he wants to do. For example, he can ask the platform to retrieve all the documents from a data room, extract certain information, synthesize them according to a specific format, then produce a final document. These workflows can then be shared within the firm, but also with clients, in order to facilitate collaboration on a file.
Do you already have clients in France and are you considering opening an office there?
We are already working with several firms in France, including Bredin Prat since 2024. This firm has since deployed our platform to more than 200 lawyers. We also have some large French companies among our clients, which we prefer not to name for the moment. Although we do not yet have a dedicated team on site, we have already hired several French lawyers within our commercial teams. The opening of an office cannot be ruled out but we have nothing planned at this time. In each country where Harvey expands, our method involves first collaborating with local law firms, before building the necessary legal databases and recruiting local teams.
Harvey has raised $760 million in 2025, including $160 million in December at a valuation of $8 billion, led by Andreessen Horowitz. In August, the company generated nearly 100 million dollars in ARR, according to TechCrunch. How do you explain this enthusiasm among investors? ?
In my opinion, generative AI represents a major technological breakthrough, comparable to the arrival of the personal computer or the Internet. We have successfully demonstrated a clear growth trajectory, strong adoption by leading customers and a market opportunity validated by our traction. Harvey claims more than 700 customers in 58 countries. Our company was able to raise funds from renowned investors such as Kleiner Perkins, Sequoia, Andressen Horowitz, EQT, etc.
How will generative AI transform the legal profession in the years to come?
It’s complicated to predict. I think the profession will evolve, just like programming where the way of coding is changing. My belief is that lawyers will become more and more valuable, and that the best ones will be able, like the best programmers, to accomplish more because AI will give them more leverage. Some simple tasks will likely become less onerous, such as writing an NDA that AI models can perform. On the other hand, complex and important legal subjects will always be handled by expert and specialized humans. AI models do not replace this.
Despite the rapid progress of AI, you therefore believe that the profession of lawyer will remain fundamentally irreplaceable ?
When you hire a large firm for a major dispute or an important transaction that puts the life of your business at stake, you are paying for the commitment and competence of that firm. This is where the value lies. I think customers will always want to have humans in the loop, for performance reasons but also regulatory reasons. So I don’t believe at all that the legal profession will disappear, just as I don’t believe that we will have software applications entirely designed by AI models in the next three to five years, without humans to improve or test them.
Gabe Pereyra is president and co-founder of Harvey, an AI platform dedicated to professional legal services, including law, tax and finance. Before founding Harvey in 2022, he was an artificial intelligence researcher at Meta and Google. A graduate in computer science from the University of Southern California, he also undertook a doctorate in neuroscience at the University of Oxford, funded by DeepMind, before devoting himself to entrepreneurship.
