Harvey AI: Enterprise Legal Platform Evaluation

What Harvey Is — and What It Is Not

Harvey is an AI platform purpose-built for legal work, developed by Harvey AI (formerly Harvey Intelligence) and backed by a series of high-profile venture rounds. It is not a general-purpose chatbot with a legal skin on it. The platform runs on large language models fine-tuned on legal corpora and deployed through enterprise contracts with specific data isolation commitments — a distinction that matters when evaluating it against tools like Lexis+ AI or Westlaw CoCounsel.

Harvey's primary design target is the large law firm associate and the in-house legal team handling high-volume, document-intensive work. It is not designed for solo practitioners or small firms — the pricing model and onboarding structure both reflect an enterprise-first orientation. Buyers evaluating it for a boutique or regional firm should be clear-eyed about that fit problem before engaging.

Declared Use Cases

Harvey publicly positions the platform across four primary workflow categories. Each carries different reliability characteristics that practitioners should understand before deploying:

• Legal research and memo drafting — Harvey can draft research memos, summarize case law, and synthesize statutory frameworks. This is the use case with the most documented deployment evidence at large firms. Citation accuracy is the primary risk factor; see the limitations section below.
• Contract review and redlining — The platform can identify non-standard clauses, flag deviations from a firm's preferred positions, and generate redline comments. Performance varies significantly by contract type; complex multi-party agreements require more human review than standard commercial contracts.
• Regulatory and compliance analysis — Harvey can map regulatory requirements to document provisions and flag compliance gaps. In-house teams at financial services and life sciences companies are documented users. This use case benefits most from firm-specific fine-tuning.
• Due diligence — Document triage, issue extraction, and summary generation across large data rooms. Several Am Law 100 firms have publicly disclosed using Harvey for M&A due diligence acceleration.

Harvey also supports deposition preparation and litigation drafting through its general document generation capabilities, though these are not positioned as distinct product modules — they fall under the general drafting and research interface.

Data Privacy and Isolation Model

This is the area where Harvey's enterprise positioning is most clearly differentiated from consumer-adjacent tools. Harvey's standard enterprise agreement includes a zero-retention commitment: client queries and documents are not used to train shared models. Each firm deployment operates in an isolated environment.

Harvey's underlying model infrastructure runs on OpenAI's API with enterprise data processing agreements in place. This means client data does not flow through OpenAI's consumer products, but it does mean the processing chain involves a third-party API provider — a fact that some firms with heightened confidentiality obligations (government contractors, firms handling classified-adjacent matters) have flagged as a consideration. Harvey offers private cloud deployment options for firms with stricter requirements, though pricing and availability for that configuration are negotiated individually.

Pricing Structure

Harvey does not publish list pricing. Contracts are negotiated enterprise agreements, typically structured around seat-based licensing with a minimum commitment threshold. Based on publicly reported figures and disclosed firm deployments, entry-level enterprise agreements have been reported in the range of $50,000–$150,000 annually for smaller deployments, with large firm contracts at Am Law 100 scale running substantially higher.

Firm Deployments and Evidence Base

Harvey has disclosed partnerships or deployments with a number of named firms, including Allen & Overy (now A&O Shearman), PwC Legal, and several Am Law 50 firms. These are among the most publicly documented enterprise legal AI deployments anywhere in the market, which gives Harvey a more verifiable deployment record than many competitors.

The A&O Shearman deployment is particularly well-documented: the firm rolled out Harvey to attorneys across multiple practice groups and geographies, and has published internal productivity metrics (time saved on research tasks, memo drafting speed). Those figures should be read as firm-reported, not independently audited — but the deployment scale itself is verifiable through public announcements.

Citation Accuracy and Known Limitations

Citation reliability is the most consequential accuracy dimension for legal research tools, and it is where Harvey's limitations require the most attention from buyers.

Harvey uses retrieval-augmented generation (RAG) for legal research tasks, grounding outputs in retrieved documents rather than relying purely on model weights. This reduces — but does not eliminate — hallucination risk. The platform does not have direct integration with Westlaw or LexisNexis databases; it relies on its own legal corpus and, in some configurations, firm-provided document libraries.

Documented Limitation Areas

• No native Westlaw/Lexis integration: Research outputs require separate verification against a licensed legal database. This adds a step that integrated research tools eliminate.
• Jurisdiction coverage gaps: Performance is strongest on US federal and major state law. Coverage of non-US jurisdictions, particularly for civil law systems, is uneven. Harvey has expanded international coverage through its A&O Shearman deployment, but practitioners in non-common-law jurisdictions should test before committing.
• Regulatory currency: The training corpus has a knowledge cutoff, and regulatory analysis outputs may not reflect recent agency guidance or rulemaking. This is particularly relevant for fast-moving areas like AI regulation, securities enforcement, and environmental compliance.
• Complex multi-party contracts: Contract review quality degrades on highly negotiated, multi-party agreements with cross-references and complex defined term networks. Standard commercial contracts perform substantially better.

How Harvey Compares to Directly Competing Platforms

The table below summarizes Harvey's positioning against Westlaw CoCounsel and Lexis+ AI on the dimensions most relevant to large firm and in-house buyers. These are the three platforms most frequently evaluated head-to-head in enterprise procurement processes.

The most practically significant difference: if a firm's primary use case is legal research and they already subscribe to Westlaw or Lexis, CoCounsel and Lexis+ AI have a structural advantage because citations are live and verifiable within the same platform. Harvey's advantage is in drafting-heavy workflows — due diligence, regulatory analysis, contract generation — where the absence of native database integration matters less.

Target Audience: Who Harvey Is a Good Fit For

Strong fit

• Am Law 100–200 firms with high-volume M&A, private equity, or regulatory practices where drafting and due diligence acceleration is the primary value driver.
• In-house legal teams at large enterprises — particularly financial services, life sciences, and technology companies — with significant contract volume and regulatory analysis needs.
• Firms willing to invest in custom fine-tuning to align Harvey's outputs with house style and preferred positions, and that have the IT infrastructure to support an enterprise SaaS deployment.
• International firms with US and UK practices that can leverage Harvey's strongest coverage areas while accepting more limited performance in other jurisdictions.

Poor fit

• Solo practitioners and small firms: the enterprise minimum commitment and pricing structure make Harvey economically inaccessible for practices below a certain scale.
• Litigation-heavy practices where primary-source citation verification is the dominant workflow need — integrated research tools have a structural advantage here.
• Practices operating primarily in non-US, non-UK jurisdictions where Harvey's corpus coverage is thinner and less tested.
• Firms that require on-premises deployment without a cloud component, unless they negotiate Harvey's private cloud option specifically.

Supervision and Professional Responsibility Considerations

Harvey's outputs — research memos, contract redlines, regulatory analyses — require attorney review before use in client matters. This is not a Harvey-specific observation; it applies to every generative AI tool in this category. But it is worth stating explicitly because Harvey's quality on drafting tasks is high enough that there is a real risk of under-review: outputs can look polished and authoritative while containing substantive errors that a less fluent draft would have flagged.

Supervising attorneys remain responsible for the accuracy of all work product. Several state bar ethics opinions issued through 2025 and into 2026 have addressed this point directly in the context of AI-assisted drafting, generally concluding that competence under Model Rule 1.1 requires understanding both the capabilities and the limitations of the tools being used.

Evaluation Summary

Harvey is a genuinely capable enterprise legal AI platform with a verifiable deployment record at major firms. Its strengths are in drafting-intensive workflows — due diligence, regulatory analysis, contract review — where it can meaningfully reduce time on first-draft work. Its limitations are real: no native integration with primary legal databases, uneven international coverage, and a pricing model that excludes most of the legal market by design.

For large firm and in-house buyers evaluating Harvey, the honest procurement question is not whether it is a good product — it is whether the specific workflow mix justifies the enterprise investment, and whether the absence of native citation linking creates an acceptable or unacceptable verification overhead for the practice groups that will use it most.