Overview
Harvey is purpose-built to help legal professionals streamline workflows, analyze vast amounts of data, and generate precise, actionable insights. Our platform is engineered for mastery—unlocking a faster flow state so you can focus on high-impact work.
Harvey’s core tools—Assistant, Vault, Workflows, History, and Library—enable you to automate routine tasks, extract key insights, and generate complex legal documents with speed and accuracy.
Harvey’s Core Tools
Assistant: Ask, Analyze, Draft
Harvey’s Assistant is designed for two key functions:
Assist: Best for quick responses, all-purpose summarization, brainstorming, comparison tables/charts, and follow-up Q&A.
Tip: Upload documents or select a knowledge source like EDGAR to search live SEC fillings. Harvey will generate responses with citations from attached documentation or vetted sources.
Draft: Generates longer, structured legal documents, including NDAs, memoranda, contracts, and brief sections. Once a draft is produced, edit the output in a tracked changes format.
Learn more about Assistant.
Vault: Scale Document Review & Analysis
Vault enables professionals to upload, store, and analyze thousands of documents efficiently. It’s designed for large-scale due diligence, contract analysis, and litigation review.
Review Tables: Automatically organizes key data points into a structured table.
Targeted Q&A: Extract precise information from vast datasets using Assistant or Draft.
Bulk Summarization: Quickly process and compare large volumes of contracts or legal filings.
Learn more about Vault.
Workflows: Automate & Optimize Legal Tasks
Choose from a variety of workflows to expedite and streamline routine tasks.
Analyze a Redline Document: Automatically generates custom issue lists and key change analysis.
Translate Documents: Convert Word or PDF files across multiple languages. (Note: This is not a substitute for certified translation services.)
Vault Workflows: Pre-built tables for reviewing document sets with zero setup required.
Learn more about Workflows.
History: Track & Retrieve Your Work
Access past queries, responses, and documents.
Filter by keywords, date, client, or document count.
Save and organize frequently used queries for future reference.
Learn more about History.
Library: Curated Prompts & Examples
Harvey’s Library is designed for learning and efficiency:
Prompts: Save frequently used queries and best-practice templates.
Examples: Review successful queries and outputs curated by Harvey and your team.
Learn more about Library.
Optimizing Your Queries
Be Explicit: Open-ended questions increase the risk of hallucinations. Instead of “What’s the law on financial reporting?”, specify “Summarize Delaware’s financial reporting obligations for public companies.”
Provide Context: Describe the relevant facts, governing law, or uploaded documents.
Use Proper and Clear Formatting: Request responses in tables, bullet points, or structured summaries for precision and always include punctuation at the end of your query.
Break Down Complex Tasks: Ask one specific question per sentence for best results.
Define Your Audience & Tone:
“Summarize for a senior in-house counsel.”
“Draft in a neutral, formal tone.”
Learn more about Prompt Writing.
Reducing Hallucinations & Improving Accuracy
Harvey is engineered to minimize hallucinations, but human verification is critical.
Best Practices for Reliable Outputs:
Upload Supporting Documents or Use a Knowledge Source: Ground Harvey’s responses in vetted legal sources or relevant documents instead of relying on open-ended research questions.
Refine & Iterate: Adjust queries based on responses to improve accuracy.
Avoid Requesting Citations Explicitly: Harvey will automatically generate citations when appropriate. Requesting specific citation formats or asking for references unnecessarily may reduce output quality.
Learn more in our Harvey AI FAQ.
Additional Considerations
Harvey’s Training Data: Harvey is trained on legal knowledge but does not have real-time internet access. Use uploaded documents or knowledge sources for the most up-to-date insights and be mindful of Harvey's 'knowledge cut-off' dates.
Answers May Vary: Harvey's responses may differ based on slight changes in query wording or repeated submissions. Refine your query to explore different perspectives or improve precision.
Harvey Does Not Perform Calculations: While it can retrieve numerical data, it does not execute mathematical computations.
If you do want to experiment with Harvey's ability to calculate, ask for the components of the calculation first, then ask for the specific calculation.
Harvey Does Not Process Images: Ensure all relevant data is included in text format.
Word Count Flexibility: Instead of requesting a “500-word summary,” specify “brief,” “detailed,” or “in 10 bullet points.”
Assistant Does Not Detect Document Inconsistencies: Assistant is optimized for extracting relevant information rather than cross-referencing discrepancies between sources. To analyze changes between two document versions, use the Analyze a Redline Document workflow.