AI for Legal

Why AI matters in legal

Done well, AI in legal cuts associate hours on first drafts by 40-70%, accelerates due diligence from weeks to days, and surfaces contract risk faster than humans. Done badly, it hallucinates cases, exposes client data, and creates malpractice risk. The line between great and terrible is governance.

The core legal AI stack

Legal-specific

• Harvey — leader for elite firms.
• Casetext CoCounsel — Thomson Reuters' offering, broad availability.
• Spellbook, Lawgeex — contract review.
• NetDocuments, iManage — DMS with AI features.

General-purpose with guardrails

• Claude with enterprise zero-data-retention — strong reasoning, careful with citations.
• Microsoft Copilot — for Word + Outlook integration.

Deep dive 1: Drafting and first-pass review

The workflow: associate uploads precedent + redline → AI proposes structure → drafts clauses → associate reviews and refines.

Where it shines: NDAs, standard commercial agreements, employment agreements, vendor contracts. Where it doesn't: novel matters, jurisdiction-specific regulatory work, anything requiring extensive case law citation.

Deep dive 2: Due diligence and document review

AI tags documents in a data room, extracts key terms, flags anomalies, summarizes risk. Cuts a 3-week diligence to 3 days.

The control: senior associate spot-checks at least 10% of AI-tagged docs. Don't trust the tags without verification.

Deep dive 3: Contract analysis at scale

In-house teams: parse your contract book for renewal dates, MFN clauses, change-of-control provisions, unusual terms. Build a queryable database.

Governance

• Client confidentiality. PHI/financial/privileged content never in public LLMs.
• No hallucinated citations. Always verify cases the AI cites — courts are sanctioning attorneys for fabricated citations.
• Conflicts. AI tools that learn across matters can create conflict risks. Use vendor-isolated environments.
• Privilege. Document AI use in matter files to preserve work-product protection.
• Bar opinions. Most state bars have issued guidance. Follow it.

30-60-90 day plan

Days 1-30: select tool (Harvey or Casetext). Pilot on one practice area. BAA + DPA in place.

Days 31-60: roll out to a second practice area. Build internal training. Document AI use policy.

Days 61-90: measure: associate hours saved, time-to-first-draft, partner satisfaction with output quality. Iterate.

Maturity model

• Level 1: Partners use ChatGPT for brainstorming behind closed doors.
• Level 2: Approved AI tool deployed firm-wide. Drafting + DD + contract review pilots.
• Level 3: AI baked into matter workflow. Time savings tracked and billed appropriately.
• Level 4: Practice-area-specific AI agents handling repetitive matters end-to-end.

Where to go next

• Browse Legal tools → (filter to Legal)
• For matter intake automation, see the AI Chief of Staff playbook.