Harvey, a developer of artificial intelligence software for legal professionals, today announced that it has raised $300 million in a new late-stage funding. 

The Series E investment was led by Kleiner Perkins and Coatue. They were joined by more than a half-dozen other backers, including Alphabet Inc.’s GV investment arm, the Open AI Startup Fund and Sequoia. The funding round values Harvey at $5 billion, up from $2 billion following its previous raise earlier this year.

Harvey, officially Counsel AI, develops AI tools that promise to make legal teams more productive. The company says that its platform is used by more than 330 law firms and other organizations in 50-plus countries. Harvey reportedly generated $75 million in annualized recurring revenue from those customers as of April, up from $50 million a few months earlier.

Attorneys often start the process of drafting a new legal document by researching relevant laws and court decisions. Harvey’s flagship AI tool, Harvey Assistant, can automate the task. It enables legal teams to aggregate information from court filings and other sources using natural language prompts. The tool includes citations in its prompt responses to help legal teams verify the accuracy of the output.

After collecting the documents necessary for a project, Harvey Assistant can automate the document drafting process. From there, it checks the draft against an organization’s internal guidelines for legal filings. It can also proofread newly created documents, translate them to other languages and, in the case of contracts, autofill the required information into each field.

Complex legal agreements often go through dozens of revisions before they’re finalized. Attorneys indicate that a section should be deleted or modified with a red strikethrough, which is why the editing process is commonly known as redlining. Harvey says that its platform can automatically summarize such document changes to streamline the revision management process for users.

The company built a custom AI model to power its redlining capabilities. It features about 100,000 parameters and uses a convoluted neural network architecture, which is commonly used for image processing tasks. 

Before inputting a contract into the AI model, Harvey Assistant turns the PDF file that contains it into an image. The software then applies a grayscale filter. According to the company, its engineers have determined that this arrangement increases the AI model’s accuracy and speeds up inference.

Harvey’s other features are powered by off-the-shelf large language models from providers such OpenAI. Going forward, it also plans to use AI models from LexisNexis, a provider of legal document databases and software tools for attorneys. LexisNexis parent RELX plc participated in Harvey’s newly announced funding round through its REV venture capital arm.

According to Fortune, Harvey will use the proceeds from the investment to double its more than 340-strong workforce. The company also plans to expand into new market segments beyond the legal profession.

Image: Pixabay