What Harvey does
Harvey is an enterprise AI platform for legal and professional-services teams to research, analyze documents, draft work product, and build governed workflows.
Harvey is designed as an enterprise work platform rather than a consumer legal chatbot. Lawyers and other professionals can ask questions, analyze large document sets, draft and revise material, and configure repeatable workflows around their own knowledge. Its orchestration layer can route work across different models and regions, while firm content and matter context make outputs more specific than a general prompt. The practical value is a shared environment for moving from source material to an editable first pass without scattering confidential documents across unrelated tools.
Harvey does not publish a standard subscription price. Prospective customers request a demo and negotiated enterprise quote, so total cost depends on users, products, integrations, support, region, implementation, retention, and contractual controls. Procurement should require a precise entitlement map and a representative pilot rather than extrapolating from vendor case studies. Compare time saved only after counting source preparation, prompt design, professional verification, corrections, knowledge maintenance, and change management. A fast draft is not valuable if a senior lawyer must reconstruct its authorities or assumptions.
Legal work makes confidentiality and accuracy inseparable. Harvey says customer documents, prompts, and responses are not used to train its models or underlying providers, requires zero data retention from model providers, and offers SAML SSO, audit logs, IP allow-listing, lifecycle controls, and regional processing. It lists SOC 2 Type II and ISO 27001, 27701, and 42001. Those controls do not validate an answer. Lawyers must check authorities, quotations, jurisdiction, dates, privilege, conflicts, client instructions, and final reasoning before advice, filing, negotiation, or disclosure.
How Harvey works
A professional supplies a question, instruction, document set, or approved knowledge source. Harvey orchestrates an appropriate model and workflow to research, extract, compare, reason, or draft, then returns work product and supporting context for professional verification. Workspace permissions, regional processing, retention, and integrations govern which material may enter each matter.
Establish the authorized context
Start with a defined legal task, jurisdiction, client instruction, approved documents, confidentiality classification, and expected deliverable. Matter separation and permissions determine which knowledge Harvey may retrieve.
Route work through models and tools
Harvey interprets the request and applies an appropriate model, research path, document analysis, or configured workflow. Regional and model settings govern processing, while source quality controls the available evidence.
Generate an evidence-linked first pass
The platform returns research, extracted facts, comparisons, analysis, or draft language for editing. Users should require traceable support and clearly distinguish source text, inference, assumption, and proposed professional judgment.
Exercise professional responsibility
A qualified reviewer opens authorities, checks jurisdiction and currency, tests quotations and reasoning, protects privilege, and approves the final work. Corrections should feed workflow and knowledge maintenance rather than disappear.
How to set up Harvey
Define approved matters and tasks
Set practice areas, jurisdictions, client-consent requirements, excluded data, privilege rules, retention, owners, and the exact work product a pilot may create.
Complete security and contract review
Verify region, subprocessors, model routes, zero-retention terms, certifications, incident obligations, deletion, audit access, integrations, support, and the full enterprise quote.
Configure identity and knowledge
Enable SSO, roles, workspaces, matter separation, retention, and least-privilege repositories; upload only current, authorized sources with clear ownership.
Benchmark representative work
Test known research questions, contract analyses, citations, adverse authorities, ambiguous instructions, restricted matters, and deliberate prompt attacks against expert answers.
Release with professional controls
Require source inspection and named lawyer approval, log material corrections, monitor access and adoption, and suspend workflows whose error or review burden exceeds value.
Harvey FAQs
How much does Harvey cost?
Harvey does not publish a standard rate card. Firms request a demo and enterprise quote based on products, users, region, integrations, implementation, and support.
Does Harvey provide legal advice?
Harvey assists qualified professionals with research, analysis, and drafting. Its output is not a substitute for a lawyer's independent judgment or client-specific advice.
Does Harvey train on customer documents or prompts?
Harvey says it contractually does not use customer data or content to train models and requires zero data retention from underlying model providers.
Can Harvey process data in different regions?
Harvey documents regional options including the US, EU and Switzerland, and Australia. Buyers should confirm the exact feature, model, backup, and subprocessor data flow in contract.
Why must citations still be checked?
AI can cite the wrong authority, misstate a holding, miss adverse law, or apply the wrong jurisdiction. A professional must open and validate every material source.
Listing reviewed 2026-07-15. Product details and pricing can change; verify important terms on the provider's website.
Related Research AI tools
Related AI guides
Reviews
Tell the community what you made, what worked, and what you wish you knew before starting.