Harvey
Enterprise legal AI for high-stakes research, drafting, and review
About
Harvey is an AI platform built specifically for legal and professional services organizations, focusing on workflows for leading law firms and corporate legal teams worldwide. It connects to a firm’s knowledge base and, in many deployments, to premium legal content providers, enabling lawyers to query, summarize, and draft against both public legal materials and proprietary documents within a controlled environment. The platform is designed to slot into existing legal workflows rather than operate as a general-purpose chatbot, with guardrails and auditing features that support professional standards. Functionally, Harvey supports core legal tasks such as contract analysis, due diligence, compliance review, and legal research. Lawyers can use it to generate first drafts of contracts, pleadings, memos, and emails, to extract structured data from large document sets, and to answer complex legal questions grounded in a firm’s own knowledge and licensed content. Advanced retrieval and reasoning capabilities help it surface relevant clauses, cases, and regulations at scale, turning previously manual review exercises into more targeted, high-value analysis for attorneys. The platform is delivered as enterprise software with a strong focus on security, privacy, and governance appropriate for Am Law 100 firms and Fortune 500 legal departments. It typically integrates with document management systems and identity providers so firms can enforce role-based access, matter permissions, and audit logging. Harvey’s internal tooling includes analytics and ROI calculators that help firms estimate time savings and profit impact from deploying legal AI across practice groups. A key differentiator for Harvey is its concentration on sophisticated, high-stakes legal work rather than general business use. Industry analyses and commentary describe it as aimed at lawyers billing high hourly rates and needing AI support on complex tasks like multi-jurisdictional research, large-scale M&A due diligence, and intricate contract portfolios. This focus, together with deep integrations into legal content and firm systems, positions Harvey as a premium, enterprise-grade legal AI solution rather than a mass-market productivity tool.
What you can do with it
- Generate first-draft contracts, pleadings, and legal memos based on firm templates and governing law
- Conduct large-scale M&A due diligence by summarizing and flagging issues across thousands of deal documents
- Analyze high-volume NDAs and commercial contracts to identify non-standard clauses and compliance risks
- Research and summarize case law and statutes across multiple jurisdictions for a specific legal question
- Assist in-house legal teams with rapid review and redlining of vendor and customer agreements
Pricing
Unconfirmed
How to access
Harvey is accessed primarily via a secure web application deployed for law firms and corporate legal teams, with access provisioned through enterprise sales rather than open signup; firms typically integrate it with their document management systems and single sign-on providers, and users work within the browser-based interface to run analyses, drafts, and research workflows.
Access is via the Harvey web app with authenticated accounts provisioned for law firms and corporate legal departments; there is no self-serve signup and prospective customers are directed to book a demo or contact sales for enterprise access, typically using SSO or firm-managed identity providers.
Tips for getting the best results
To use Harvey effectively, firms first complete an enterprise onboarding process where identity, permissions, and document system integrations are configured; this ensures matters and practice groups have appropriate access controls. Lawyers then interact with Harvey through structured workflows for tasks like contract review, due diligence, and research, typically starting from curated templates or prompt patterns designed for those workflows. Users are encouraged to provide detailed, jurisdiction-specific instructions, include relevant governing law and matter context, and, where possible, attach or reference specific documents to ground the model’s responses. Outputs should be treated as high-quality drafts or research starting points and systematically checked against primary sources and firm standards, particularly for novel issues or unfamiliar jurisdictions.
Known limitations
Harvey is an enterprise-only platform with no public, self-serve access, making it effectively unavailable to solo practitioners and small firms. Public pricing is absent, and independent analyses describe it as a premium product aimed at top-end legal teams, so costs and minimum commitments may be prohibitive for many organizations. As with all generative AI, its outputs can contain errors, omissions, or hallucinations and must be reviewed by qualified counsel, especially for nuanced or cutting-edge legal questions. Its strongest use cases are in supported jurisdictions and practice areas; coverage and performance may be more limited where underlying content or training data are sparse. Because deployments depend on firm integrations and configurations, realizing full value requires IT involvement, change management, and careful governance rather than plug-and-play use.
Model / Technology
Proprietary large language models and RAG pipeline over legal content and firm documents
Commercial use
Harvey’s legal terms emphasize professional and enterprise use, with outputs intended to assist but not replace lawyer judgment; firms retain responsibility for review and compliance, and any specific commercial licensing or redistribution rights for outputs are governed by confidential customer agreements rather than public, standardized terms.
Training data
Public information indicates Harvey is trained on a mix of licensed legal content, public legal materials such as case law, and customer-provided documents under strict confidentiality, with additional access in some deployments to proprietary databases like LexisNexis through negotiated partnerships; precise training corpora, data mixtures, and any web-scale components are not publicly detailed and remain largely confidential.