Files
Review generated DD artifacts and uploaded project documents in one place, without leaving the console shell.
Recent files
25 items in the current scope.
Parsor · Screening Card
Parsor
Parsor · IC Memo
Parsor
Perplexity Ai · IC Memo
Perplexity AI
Perplexity Ai · Screening Card
Perplexity AI
Merchant Payments - Enterpise Crypto Banking v6.pdf
Merchant Payments Enterpise Crypto Banking V6
Dune Polygala Ai · IC Memo
Dune (polygala-ai)
Dune Polygala Ai · Screening Card
Dune (polygala-ai)
Dune By Polygala Ai · IC Memo
Dune (by polygala-ai)
Dune By Polygala Ai · Screening Card
Dune (by polygala-ai)
Paperclip Ai Orchestration · Screening Card
Paperclip (AI Orchestration)
Paperclip Ai Orchestration · IC Memo
Paperclip (AI Orchestration)
Connectonion · Screening Card
ConnectOnion
Cursor Anysphere · Screening Card
Cursor (Anysphere)
Cursor Anysphere · IC Memo
Cursor (Anysphere)
Legora · Screening Card
Legora
Legora · Screening Card
Legora
Harvey · Screening Card
Harvey
Harvey · Screening Card
Harvey
Lovable · Screening Card
Lovable
Lovable · Screening Card
Lovable
Orthogonal · Screening Card
Orthogonal
Aiyu Intelligence · Screening Card
Aiyu Intelligence
Taku · Screening Card
Taku
Spatial Walk · Screening Card
Spatial Walk
Spatial Walk · Screening Card
Spatial Walk
Harvey · Screening Card
archive/dd-reports/harvey_ic_memo.md
Type
Screening Card
Subject
Harvey
Updated
3月23日 17:42
Size
5.6 KB
Preview
Rendered directly from the stored file.
⚠️ This due diligence report is generated from publicly available information and AI analysis. It should supplement, not replace, human judgment, reference calls, and direct founder interactions.
Investment Committee Memo: Harvey
Date: 2026-03-23 Company: Harvey Sector: LegalTech / Enterprise AI Stage: Growth ($11B Valuation / $195M ARR) Recommendation: PROCEED WITH CAUTION (DEEPER LOOK REQUIRED)
I. Executive Summary
Harvey has rapidly established itself as the premier AI platform for "Big Law," capturing nearly 50% of the Am Law 100 and securing major strategic partnerships with PwC and A&O Shearman. However, as of Q1 2026, the company faces a critical "Verification Paradox": its massive $11B valuation (56x ARR) is predicated on massive efficiency gains that its current technical performance (~17% hallucination rate) cannot fully support without intense human oversight. This memo outlines the case for and against further investment, highlighting the tension between market leadership and technical reliability.
II. Company Overview
Founded by Winston Weinberg and Gabe Pereyra, Harvey provides a specialized generative AI platform for the professional services industry. Initially focused on legal research, drafting, and document analysis, the company is now expanding into tax, audit, and broad professional services automation.
III. Team
- Leadership: CEO Winston Weinberg (Ex-O'Melveny & Myers) and Gabe Pereyra (Ex-DeepMind/OpenAI) provide a rare blend of domain expertise and deep AI research pedigree.
- Execution Bench: Recent hires from high-growth SaaS icons like HubSpot and Figma (Tier B) suggest a maturing executive layer focused on scaling operations and sales beyond the founder-led growth phase.
- Integrity: No significant red flags; the team demonstrated increased transparency following the 2025 "Reddit Leak" by providing verifiable internal metrics.
IV. Market Opportunity
- Current Traction: $195M ARR with high penetration in top-tier legal firms.
- Expansion Strategy: Strategic move into the "Professional Services OS" market via Tax/Audit partnerships (e.g., PwC).
- TAM: Extending beyond the $30B legal software market into the $600B+ global professional services market.
V. Product & Technology
- Performance Gap: Independent benchmarks from Stanford/Vals AI (2025-26) show a 16.17% hallucination rate (Tier A). This creates a "Verification Paradox" where lawyers must manually audit outputs, potentially negating the promised ROI.
- Architecture: Moving from a "Chat" interface to "Matter-Centric" workflow automation. Leveraging RAG (Retrieval-Augmented Generation) grounded in licensed LexisNexis data and multi-model routing (GPT-o1/o3).
- Compliance: Building automated human-oversight logging to address the "High-Risk" designation under the upcoming EU AI Act (Aug 2026).
VI. Competitive Landscape
- Incumbent Advantage: Thomson Reuters (CoCounsel) and LexisNexis own the primary data sources Harvey must license. This creates a structural "Data Moat" risk.
- Market Dynamics: Incumbents are now bundling AI tools for "free" or low cost into existing Westlaw/Lexis subscriptions, putting pressure on Harvey’s $1,200/user/month premium pricing.
VII. Traction & Financial Overview
- Valuation: $11B (implied 56x ARR multiple), which is significantly higher than the current average for late-stage SaaS (10-15x).
- Utilization: Internal claims of 77% utilization are contested by industry surveys suggesting figures as low as 30% (Tier B). High risk of "shelfware" churn if adoption doesn't cross the "Innovation Lab" chasm.
VIII. Investment Merits (Bull Case)
- Winner-Take-Most: In a prestige-driven industry like Big Law, Harvey is the "Safe Choice" for innovation partners.
- Platform Shift: If Harvey successfully transitions from a tool to an OS for professional services, its TAM and valuation floor rise dramatically.
- Data Flywheel: Exclusive partnerships and fine-tuning on proprietary client data could eventually create a moat that incumbents cannot replicate.
IX. Risk Factors & Mitigants
- The Billable Hour Paradox: Legal firm business models structurally disincentivize efficiency gains. Mitigant: Shifting firms toward value-based pricing enabled by AI.
- EU AI Act Liability: High compliance costs and potential 7% turnover penalties. Mitigant: Rigorous SOC2/ISO 27001 and human-in-the-loop audit logs.
- The "Good Enough" Free Alternative: CoCounsel bundling. Mitigant: Superior UX and deep workflow integration that incumbents lack.
X. Valuation & Returns Analysis
At 56x ARR, Harvey is priced for perfection. A 10x return would require a ~$110B valuation, implying Harvey must become the "Microsoft of Professional Services." This requires not only holding the current market but also successfully disrupting the Tax and Audit industries while maintaining high margins despite data licensing costs.
XI. Recommendation & Next Steps
Recommendation: Proceed to deeper technical and customer due diligence.
Next Steps:
- Customer Reference Calls: Focus on "Power Users" within Am Law 100 to verify actual time-saved vs. verification-time.
- Technical Audit: Review the roadmap for hallucination reduction and RAG accuracy.
- Licensing Review: Analyze the durability and cost-escalation clauses in the LexisNexis/TR data sharing agreements.
Sources:
- Stanford HAI: Legal AI Hallucination Study (2025/26) (Tier A)
- EU AI Act Annex III - High-Risk AI Categories (Tier S)
- Sacra Harvey Analysis (Tier C)
- Business Insider/Tooliverse (Tier B)
What belongs here
Generated docs
Screening cards, memos, and other DD outputs written by the workflow.
Project uploads
Decks, PDFs, and supporting files uploaded into Korok for context.
Daily handoff
Use this tab as the review shelf before jumping back into DD.