The Ethics And Risk Frame
AI in law firms operates under three constraints that don't apply elsewhere: privilege (the AI vendor cannot become an unauthorized recipient of client confidences), competence (the lawyer remains responsible for AI outputs), and fee reasonableness(you can't bill 6 hours of partner time for what AI did in 30 seconds).
Get these right and AI is transformative. Get them wrong and you have a bar complaint waiting to happen. Every use case below is filtered through this frame.
Client Intake
The single highest-leverage AI use case in a law firm. AI conducts the initial intake conversation in chat, gathers the facts in a structured way, runs conflict-of-interest checks, and routes to the right attorney with a pre-formatted matter summary. Time savings: 60–80% per intake.
What not to automate in intake: legal advice. The AI explicitly disclaims that it is not providing legal advice and that the attorney-client relationship begins only after engagement. The AI gathers facts; the attorney evaluates and advises.
Contract Review
AI contract review is the most mature legal AI use case. The model reads the contract, extracts key terms, flags deviations from a playbook (the firm's or client's standard positions), and produces a summary memo. Junior associate work compressed from hours to minutes.
The critical guardrail: the AI's output is a first draft reviewed by an attorney. The attorney signs the memo and the attorney is responsible for its accuracy. AI is a force multiplier, not a substitute.
- • AI vendor has a privilege-preserving agreement (no training on your data, no logging).
- • Data residency matches client expectations.
- • Access logs document who saw what, when.
- • The vendor is reviewed annually for changes to terms.
Legal Research
Legal research AI is now table stakes for serious firms. The good products ground answers in primary sources (statutes, cases, regulations) with citations. They flag where authority is split. They suggest related lines of analysis.
The bad products hallucinate cases. The infamous “fake citation” incidents of 2023 should be a permanent warning. Use only research products that ground every cite in a verifiable source — and verify before filing.
Drafting and Templating
AI drafts demand letters, motions, briefs, memos, and routine correspondence from facts the attorney provides. The attorney edits and finalizes. Time savings: 40–70% on standard documents.
The non-negotiable: AI never drafts where citations matter (briefs, motions, memos) without a hard requirement that every cite is verified against the source. Without the verification step, you have the 2023 problem again.
Billing and Time Tracking
AI converts daily activity logs (calendar entries, emails, document opens) into time-tracking suggestions. The attorney reviews and confirms. Massively reduces under-billing — most attorneys leave 10–25% of their billable time on the table because they didn't log it.
Critical: never auto-bill without attorney approval. The attorney is the only person who can ethically attest to billable time.
Compliance and Confidentiality
A law firm AI deployment must satisfy: the bar's ethics rules, state data breach laws, client engagement letter terms, and any industry-specific rules (HIPAA for health clients, etc.). Document the compliance review and re-do it annually.
Vendor Selection
Three vendor categories now exist: legal-specific AI products (Harvey, Spellbook, others), general-purpose AI with enterprise terms (OpenAI, Anthropic), and internal builds. For most firms under 50 attorneys, a legal-specific product is the right starting point — you get the privilege-preserving terms and a tuned model without building the infrastructure.
AI in law firms isn't a question of whether to adopt — it's a question of whether your firm adopts before clients start asking why their fees haven't come down.
For broader implementation see our SMB roadmap and our AI systems service.
FAQ
What about hallucination risk in legal documents? Verify every citation against primary sources before filing. Non-negotiable.
Does the bar prohibit AI?No, but most state bars now have guidance. Read your state's recent ethics opinions.
How do we tell clients? Disclose AI use in engagement letters. Many clients now expect it; transparency wins.