Harvey Launches End-to-End Legal Agents at $11B Valuation

Harvey, the AI platform built for law firms and corporate legal departments, is pushing its product further up the work stack. The company today is highlighting Harvey Agents — AI workers that execute legal tasks end-to-end rather than assisting on individual steps — along with updates to its Agent Builder that let firms spin up custom agents in minutes. The company now has more than 100,000 lawyers at 1,300 organizations using the platform, and it's doing this on the back of a fresh $200 million Series F that valued the company at $11 billion.

If the legal industry's question in 2025 was "does AI assist lawyers?", the question for 2026 is "what work does AI do without one?" Harvey is placing a concrete answer on the table.

What "End-to-End" Actually Means

Most legal AI tools today are assistants. A lawyer prompts them, reviews the output, and edits. The workflow is still human-first with AI augmentation.

Harvey Agents flip that. A user describes the task — "run due diligence on this data room and flag material risks," or "draft a client memo analyzing the contract amendments against our playbook," or "redline this NDA per our standard positions" — and the agent plans the steps, pulls the relevant documents, performs the analysis, drafts the output, and hands the lawyer a finished work product to review.

The practical effect: the lawyer's role moves from doing the work to checking it. For standardized, document-heavy tasks, that's a big unlock. Due diligence, contract review, memo drafting, and transaction diligence are all well-suited to this pattern because they follow repeatable logic and have clear deliverables.

Custom Agents at Scale

The other notable update is Harvey's Agent Builder. Over 25,000 custom agents are now running on the platform, built by the firms themselves rather than by Harvey. That's the real signal of adoption. When your customers are building their own tools on top of your product, you've stopped being software and started being infrastructure.

Agent Builder lets users describe the knowledge sources and review tables they want embedded, and the platform handles the rest. Harvey's Vault component syncs folders from preferred document management systems (iManage, NetDocuments, SharePoint) so diligence materials, precedent libraries, and deal documents stay current without manual uploads.

A mid-sized law firm can now build a "securities offering diligence agent" tailored to its house style in an afternoon. That used to require either expensive custom software or large internal tech teams. Neither are prerequisites anymore.

Who's Using It

Harvey's footprint is no longer hypothetical. The platform serves most of the AmLaw 100 (the largest U.S. law firms by revenue), more than 500 in-house legal teams, and 50 asset management firms — the last category being significant because buy-side investment firms have historically been slower to adopt legal AI than their law firm counterparts.

More than 100,000 lawyers use Harvey across 1,300 organizations in over 60 countries. For a product that launched in 2022, that's enterprise-scale adoption in an industry often considered slow-moving.

The $11 Billion Question

Harvey's valuation jumped from $5 billion in 2024 to $11 billion with this year's Series F, led by GIC with participation from Coatue. The round brought total funding to roughly $800 million.

Why are investors willing to value a legal AI company at $11 billion? The thesis is straightforward: legal services is a $900 billion global market, lawyer hours are the main input, and AI can compress those hours by double-digit percentages on a large fraction of the work. Even modest penetration translates into a very large revenue pool.

The risk, as always with vertical AI, is that the underlying models (from Anthropic, OpenAI, Google) improve fast enough to commoditize specialized layers. Harvey's defense is the integration work — the document management hooks, the playbook support, the firm-specific trained agents, the compliance stack around client data. That's a more defensible moat than raw model performance, which can and does change every few months.

What Lawyers Should Actually Do Now

If you work in legal, the practical questions are three. First, which parts of your day are document-heavy and playbook-driven? Those are the workflows Agent Builder is designed for. Second, can your firm's document management and knowledge systems integrate cleanly with Harvey's Vault? That integration is what makes custom agents actually useful rather than demos. Third, what's your firm's position on the "lawyer reviews agent output" model versus "lawyer does the work with AI assistance"? The answer affects training, pricing, and what you tell clients.

For in-house legal teams, the calculus is often simpler. You have budget constraints that law firms don't, and agents that automate repetitive review work translate directly into cost savings on outside counsel.

What's Next

Harvey's public roadmap points to deeper integration with Microsoft Word (document editing in-place while preserving formatting) and expanded support for multilingual work, particularly important for its growing international customer base. The company has also hinted at moving into transactional execution — not just drafting but actually running closing checklists and document collection.

Expect to see competitive responses from Thomson Reuters CoCounsel, LexisNexis, and a growing roster of AI-native legal startups over the next quarter. The legal AI race is unusually concentrated at the top, and Harvey's $11 billion valuation is both a validation and a target on its back.

The bottom line: Harvey is one of the clearest examples of AI moving from assistant to operator in a high-value knowledge industry. If you work in law or advise lawyers, spending a few hours with Agent Builder this month is time well spent. The workflow patterns it introduces — describe the task, delegate to an agent, review the output — are almost certainly the template for how legal work will be staffed by 2028.