ABA + ethics ground rules
- Model Rule 1.1 (competence) — you must understand the AI tools you use
- Model Rule 1.6 (confidentiality) — no client info to non-confidentiality-respecting AI
- Model Rule 5.3 (non-lawyer assistance) — AI counts as a non-lawyer; supervision required
- State-specific opinions — CA, NY, FL, TX all have issued formal opinions; check yours
Use cases worth deploying
- Client intake automation — chat + voice qualifying intake 24/7
- Contract review + clause extraction — 60–80% time reduction on standard review
- Legal research assistance — first-pass research, not authoritative output
- Deposition / discovery review — semantic search across thousands of documents
- Document drafting from templates — engagement letters, NDAs, demand letters
- Time entry capture — AI infers billable activity from calendar + email + docs
- E-discovery — Relativity AI, DISCO AI
Best AI tools for law firms
| Tool | Use case |
|---|---|
| Harvey AI | Big-firm legal research + drafting |
| Casetext CoCounsel (Thomson Reuters) | Research, summarization, deposition prep |
| Spellbook | Contract drafting + redlining in Word |
| Lexis+ AI | Research with citation |
| Ironclad AI / Robin AI | Contract lifecycle |
| Custom Claude + RAG | Firm-specific knowledge work |
AI intake — the highest-ROI starting point
Every firm we work with has the same starting problem: leads come in 24/7, attorneys can't field them 24/7. AI intake (voice + chat) qualifies + books consultations, captures conflict-check data, sends e-signed engagement letters for routine matters. Typical lift: 2–4x conversion on inbound leads.
What to avoid
- Generic ChatGPT for legal research — hallucinated citations cost careers (see Mata v. Avianca)
- AI-only document drafting — always lawyer review before sending
- AI in jurisdictions you don't know — never trust AI on procedural rules outside your bar
- Cheap "AI legal assistant" SaaS — most are GPT wrappers without legal training
Need a firm-specific AI plan? Book a free 30-minute call.