Getting Started with Harvey

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.

Navigate with Speed: Sidebar & Global Search

To start, Harvey’s sidebar is engineered for momentum. It puts your essential tools and most recent work within immediate reach—so you can stay in flow and move fast.

You can launch Global Search to find anything—instantly. Whether you're looking for a Vault project, a past Assistant thread, or a saved workflow, search is always just a click or keystroke away.

• Mac: ⌘ + K
• Windows/Linux: Ctrl + K
• Or, click the magnifying glass icon at the top of the sidebar

Search results surface by most recent activity, so what’s active rises to the top. Workflows appear in alphabetical order for fast retrieval.

Harvey’s Core Tools

Assistant: Ask, Analyze, Draft

Assistant helps you process information quickly:

• Assist Mode: Best for quick responses, all-purpose summarization, brainstorming, comparison tables/charts, and follow-up Q&A.
  • Tip: Select EDGAR as a knowledge source to search live SEC fillings. Users can specify the filing type and Harvey will generate responses with citations from vetted sources.
• Draft: Generate structured documents like NDAs, memoranda, contracts, and brief sections.

Vault: Large-Scale Document Review

Vault is purpose-built for professionals handling high-volume, high-stakes work. Upload, store, and analyze thousands of documents—all in one place. Use the Vault Sidebar for fast access to active projects or to launch a new one without breaking stride.

Accelerate your review with tools engineered for scale:

• One-Click Workflows: Run pre-selected sets of queries instantly using pre-built workflows, allowing for faster data extraction in areas like M&A and litigation.
• Review Tables: Automatically organize extracted key data into structured tables.
• Targeted Q&A: Extract precise information from datasets.
• Bulk Summarization: Designed for large-scale analysis and review.

Workflows: Automate & Optimize Tasks

Harvey offers both standalone workflows and Vault-integrated workflows:

• 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.

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.

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.

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.”

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.

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.