⚠️ 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.

# Deal Screening Card: Harvey

**Date:** 2026-03-23
**Analyst:** DD Memo Writer
**Stage:** Series C/Growth ($11B Valuation)
**Sector:** LegalTech / Generative AI

## One-Liner
Harvey is a professional services AI platform providing LLM-powered legal research and drafting for elite law firms, currently leading the "Big Law" AI transition while battling a significant "Verification Paradox."

## Scorecard

| Dimension    | Score | Signal   |
| ------------ | ----- | -------- |
| Team         | 8/10  | 🟢       |
| Market       | 7/10  | 🟢       |
| Competition  | 4/10  | 🟡       |
| Product/Tech | 3/10  | 🔴       |
| Financials   | 5/10  | 🟡       |
| Risk         | 7.8/10| 🔴       |
| **Overall**  | 4.9/10| 🟡       |

*Scoring: 🟢 >= 7 | 🟡 4-6 | 🔴 <= 3*
*Risk score: 🟢 <= 3 (low risk) | 🔴 >= 7 (high risk)*

## Key Strengths (Top 3)
1.  **Dominant Market Share in Big Law:** Successful penetration into ~50% of the Am Law 100 and key strategic partnerships (e.g., PwC, A&O Shearman) create a formidable first-mover advantage.
2.  **Top-Tier Executive Talent:** World-class leadership bench with experience from high-growth SaaS icons like HubSpot and Figma, mitigating typical "founder-only" execution risks.
3.  **Aggressive TAM Expansion:** Pivot from legal chat to a "Professional Services OS" (Tax/Audit) provides a path to justify its premium valuation multiple.

## Key Concerns (Top 3)
1.  **The Verification Paradox:** A 16-17% hallucination rate (Tier A: Stanford study) in a "zero-error" profession forces manual verification that significantly erodes the ROI and productivity gains Harvey sells.
2.  **Data Moat Vulnerability:** Harvey remains dependent on licensing legal data from direct competitors (Thomson Reuters/LexisNexis), who are now aggressively shipping their own AI products (CoCounsel).
3.  **Regulatory & Liability Headwinds:** The upcoming EU AI Act "High-Risk" designation and recent incidents of "fake citations" (Gordon Rees 2025) create a looming liability cloud for both the startup and its users.

## Verdict
**Recommendation:** Deeper Look
**Reasoning:** Harvey is a category-defining company with impressive revenue traction, but the technical debt (hallucination rates) and the extreme valuation multiple (56x ARR) present a high-risk profile. Further diligence must confirm if the "Matter-Centric" workflow shift can solve the utilization and accuracy issues.

## Questions for Founder Meeting
1.  **Utilization Discrepancy:** Industry surveys suggest ~30% utilization while Harvey claims 77%; can you provide anonymized logs or "Matter-Centric" engagement data to clarify actual seat value?
2.  **The TR/Lexis Moat:** How does Harvey's long-term margin profile survive if Thomson Reuters or LexisNexis restricts API access or increases data licensing costs significantly?
3.  **Accuracy Roadmap:** What specific technical architecture changes (e.g., GPT-o3 integration, proprietary RAG refinements) will bring the error rate below 1% to meet EU "High-Risk" compliance standards by 2026?

---

## Sources
- **Tier S:** [EU AI Act Annex III - High-Risk AI Categories](https://artificialintelligenceact.eu/annex/3/)
- **Tier A:** [Stanford HAI: Legal AI Hallucination Study (Updated 2025/26)](https://office.qz.com/harvey-ai-hit-8-billion-its-tools-still-hallucinate-in-one-of-every-six-queries-812d64182dc4)
- **Tier B:** [Business Insider: Harvey CEO Response to Reddit Leak (Sept 2025)](https://www.businessinsider.com/harvey-reddit-drama-ceo-winston-weinberg-2025-9)
- **Tier B:** [Tooliverse: Harvey User Reviews & Watch-outs (2026)](https://tooliverse.ai/tools/harvey)
- **Tier C:** [Sacra: Harvey Valuation & Multiple Analysis](https://sacra.com/c/harvey/)