AI for law firms, without violating Model Rule 1.1.

Yes — and ABA Formal Opinion 512 (July 2024) explicitly addresses generative AI use. The core requirements: (1) Competence — attorneys must understand the technology they're using. (2) Confidentiality — AI vendors cannot train on client data; use enterprise tiers from Anthropic/OpenAI/specialized legal AI vendors that meet this requirement. (3) Reasonable fees — cannot bill traditional research time for work AI did in minutes. (4) Supervision — outputs must be reviewed by licensed attorney before delivery. Most state bars (Florida, California, New York) have issued similar guidance. The technology is permitted; the workflow design has to respect the ethics rules.

It restructures their work, not eliminates it. AI handles first-pass research, document drafts, intake summaries — what junior associates spent hours on. The work that's left for humans is higher-value: reviewing AI output, exercising judgment on ambiguous matters, handling client relationships, performing the specific tasks that require licensed-attorney accountability. Most firms we've worked with end up with leaner support staff doing higher-value work, not no support staff.

A solo or 2-5 attorney firm typically spends $25,000-$50,000 on a full AI implementation (intake + document drafting + legal research + client communication + conflict checking). Plus $300-$1,500/month in AI APIs and specialized tools (Lexis+ AI at $80-$150/seat, Westlaw Precision AI at similar). Most firms recoup the build in 90-180 days through reclaimed senior attorney time. Pilot starts at $5,000-$15,000.

On the enterprise tiers — yes, both Anthropic and OpenAI meet attorney-client confidentiality requirements. Claude API and ChatGPT API both default to no-training-on-your-data. Both have SOC 2 Type II. Both offer BAA-equivalent agreements for healthcare and similar regulated contexts. The free tiers (consumer ChatGPT, Claude.ai free plan) have weaker guarantees and are inappropriate for actual client matters. Always run client work on enterprise-tier API access. We architect this distinction into every law firm engagement.

More than 40 federal judges have issued standing orders requiring AI disclosure or banning AI in submitted documents. Check your jurisdiction. Many courts now require a certification that any AI-assisted text was reviewed and verified by counsel. The high-profile failures (Mata v. Avianca, fabricated case citations) all came from skipping the human verification step. AI is allowed in 2026 federal practice; failing to verify citations is the violation, not using AI.

Two paths. (1) Use Lexis+ AI or Westlaw Precision AI — they're built specifically for legal research and have access to case law, statutes, and secondary sources. They cost $80-$150/seat/month and integrate with your existing research tools. (2) Use general-purpose AI (Claude, GPT) connected to your firm's prior memoranda and a curated case database. Cheaper but requires more verification. Most firms we work with use both — specialized legal AI for primary research, general AI for synthesis and writing.

Florida (Op. 24-1), California (CalBar Guidelines), New York, Illinois, and Texas have all issued specific generative AI guidance for licensed attorneys. The themes are consistent: disclose to clients, supervise outputs, never bill for AI-saved time as if it were human time, never delegate professional judgment. The specific implementation differs slightly by state. We work with your firm's GC or risk-management lead to ensure local compliance during implementation.

Yes — and well. AI document review for e-discovery is one of the most mature applications of legal AI. Tools like Reveal, Relativity aiR, and DISCO Cecilia handle privilege review, responsiveness coding, and key-document identification at significantly higher accuracy than traditional linear review. Cost ranges from $0.10-$2 per document. For most small firms outside complex litigation, you don't deploy your own — you use the tools your discovery vendor offers. We can recommend vendors during the AI Readiness Assessment.

The risk isn't using AI — it's using AI badly. Specifically: (1) Failing to verify citations (the Mata v. Avianca pattern). (2) Inputting confidential client data into non-enterprise AI tools. (3) Failing to disclose to clients when their matter materially involved AI. (4) Billing as if a human did work AI did. (5) Allowing AI to give substantive legal advice to clients without attorney review. We architect implementations to avoid all five. The malpractice insurance carriers we've talked to are increasingly comfortable with disclosed, properly-supervised AI use.

Three options. (1) Book a 30-min intro call via /contact — we'll talk through your practice areas and identify the highest-ROI starting point. (2) Book the AI Readiness Assessment ($1,500) — two-hour session + written 48-hour roadmap with workflow priorities and ABA-compliance review. (3) If you've already identified what you want (typically intake triage or document drafting), send us a scope and we'll quote a fixed-price pilot within 48 hours. Most firms start with option 2.