- Never paste client information or privileged material into consumer-grade ChatGPT, Claude, or Gemini. Use a legal-specific or enterprise tool with proper data handling.
- Hallucinated case law is a documented problem. Every AI-cited authority gets verified by a human before it leaves the firm.
- Intake, marketing, and contract drafting templates are the highest-leverage, lowest-risk places to start.
- Specialty legal AI tools (Harvey, CoCounsel, Lexis+ AI, Westlaw Precision with AI, Spellbook) are built for the work and worth evaluating for any firm doing real research or drafting volume.
- Legal advice, court appearances, signatures, and client communication where empathy matters stay human. AI assists with the paperwork around them.
If you run a small law firm — solo, two-attorney partnership, small boutique — you have a problem that's specific to your industry. The work is high-stakes. The volume is also high. Your billable hours are the firm's revenue, but a meaningful share of every billable day is spent on tasks that aren't really legal work: client intake, document formatting, marketing, contract templating, research support, scheduling, email triage.
AI can take a significant amount of that off your desk. It can also get you sanctioned, sued, or disbarred if you use it carelessly. This guide is about the difference.
I'll start with the part that matters most: consumer-grade AI tools — free ChatGPT, free Claude, free Gemini — are not built for legal work. They are not bound by your firm's confidentiality obligations. They have, in well-documented cases, fabricated case citations that lawyers then filed in court. Several attorneys have been formally sanctioned for it.
This does not mean AI doesn't belong in a small law firm. It means you have to be more careful than the average small business owner, and you have to choose tools that match the stakes of the work.
The two rules that govern everything else
Rule 1: Confidentiality. Client information, privileged communication, and case-specific facts do not go into consumer AI tools. Period. Use either a legal-specific platform that explicitly contracts not to train on your inputs (Harvey, Thomson Reuters CoCounsel, Lexis+ AI, Westlaw Precision with AI, Spellbook, Clio Duo) or an enterprise version of a major AI tool with appropriate data agreements (ChatGPT Enterprise, Claude for Enterprise, Microsoft Copilot enterprise tier).
Rule 2: Verification. Every citation, every quoted authority, every factual claim AI produces gets verified by a human before it leaves the firm. Always. This is not optional. The "the AI said it" defense is not a defense.
If you internalize those two rules, you can use AI productively in a small firm. If you don't, you will eventually have a bad day.
Where AI actually helps a small law firm
There are six places AI earns its keep inside a small legal practice. The first three involve no confidential information and are wide open for consumer tools today. The last three require legal-specific or enterprise-grade tools.
1. Marketing and business development (no confidential information)
Most small firms under-market because the partners don't have time and a marketing hire feels excessive. AI fixes the math.
Claude or ChatGPT can draft:
- Blog posts on common client questions ("What to do if you get served with divorce papers in Oregon," "Why your LLC operating agreement matters more than you think," "What happens at the first appearance in a DUI case")
- Newsletter content for past clients and referral sources
- LinkedIn posts establishing the firm's areas of expertise
- Website copy for practice area pages
- Email templates for common inquiry types
None of this involves a specific matter, so none of it touches confidentiality. A weekly hour with Claude, a voice doc anchored to your firm's actual writing, and a final read by the attorney before publishing — that's the workflow.
The voice doc is important here. Legal marketing copy has a tone problem: it's either overly formal in a way that pushes clients away, or it's casual in a way that erodes credibility. With your past best writing fed in as reference, AI lands closer to your firm's actual tone than to either failure mode.
A note on advertising rules: every state bar has rules about what attorneys can and cannot say in marketing. AI does not know your jurisdiction's rules. Run drafts past your bar association's advertising guidelines before publishing.
2. Client intake and qualification (use legal-specific or enterprise tools)
Intake is a paperwork tax on every new client. Forms get filled out, conflict checks get run, engagement letters get drafted, billing gets set up. Multiply that by every inquiry that comes in, and a meaningful share of the firm's time disappears into intake.
Legal-specific intake platforms (Clio Grow, Lawmatics, Lexicata) have added AI features that handle structured intake, conflict screening, and engagement letter generation. For a small firm, evaluating one of these is usually a faster path than building anything custom.
If you stay with manual intake, the rule is the same as everywhere else: client-specific information goes into your case management system, not into consumer AI tools.
3. Internal operations and templates (use carefully)
Everything that isn't about a specific client is fair game for consumer-grade AI:
- Drafting hiring posts and screening candidate emails
- Writing internal policy documents and employee handbooks
- Building training materials for new associates and staff
- Structuring CLE notes into a reference document
- Drafting vendor and operational emails
- Building checklist templates for common firm workflows
Use Claude for nuanced writing. Use ChatGPT for structured tasks. Stay away from anything case-specific.
4. Contract and document drafting from templates (use legal-specific tools)
This is the place AI has changed legal practice fastest, and the place a small firm can see the largest hours-back-per-week impact.
Specialty tools (Spellbook for transactional, ContractPodAi, Ironclad, LawGeex for review, Kira for diligence, Harvey for general drafting) are built for legal drafting. They train on legal corpora, they connect to Word or your DMS, and they understand the structure of contracts.
For a small firm doing real volume of contract drafting or review (M&A, real estate, employment, commercial), one of these tools usually pays back inside a quarter. The workflow: AI generates the first draft from a template plus the specific deal terms. The attorney reviews and finalizes. The client gets a better, faster, more consistent document at the same or lower cost.
For lower-volume firms, the calculation is different — the per-seat cost of these tools doesn't pencil if you draft three contracts a month. In that case, the better move is a clean, well-structured template library (in Word with proper styles, or in your DMS) and AI assistance on the margins.
5. Legal research (use legal-specific tools, verify every citation)
Legal research is the place consumer AI tools have done the most damage. ChatGPT and Claude have both fabricated case citations in ways that have caused real harm to attorneys who didn't verify.
The fix: use legal research platforms with AI features that are bound to real authority — Lexis+ AI, Westlaw Precision with AI, vLex Vincent. These tools generate research summaries with linked, verifiable citations from a real legal database. The attorney still verifies. But the floor is far higher than it would be with a consumer tool.
For a small firm without budget for premium research platforms, the alternative is: use ChatGPT or Claude for issue-spotting and structuring, never for citation. Treat their output as a starting point for traditional research, not as research itself.
6. Internal knowledge management (use enterprise or self-hosted tools)
Every small firm builds up institutional knowledge — past briefs, motion templates, client communication examples, training notes, CLE summaries. Most firms store this poorly. Finding a past brief on a similar issue takes hours.
AI-powered knowledge management (Glean for general, NetDocuments with AI features, iManage with AI, or a properly-configured Notion or SharePoint setup) makes that searchable. The attorney asks "do we have anything on indemnification clauses in Oregon real estate transactions?" and the system surfaces the relevant past work.
This is a higher-investment workflow than the others. It's also one of the highest-ROI ones for firms that have been operating for more than a few years and have a knowledge base worth indexing.
A starter sequence for a firm that hasn't started yet
If you're an attorney reading this and you haven't built any of this yet, here's the order I'd recommend:
Week 1 — Education and rules. Spend an hour reading your state bar's most recent ethics opinion on AI. Spend a second hour reading the firm's existing engagement letters and policies to see what needs updating once AI is in the workflow. This is unglamorous and essential.
Weeks 2–4 — Non-confidential marketing experiments. Use Claude or ChatGPT for one piece of marketing content a week. Blog post, LinkedIn post, newsletter draft. Get comfortable with the tool and build the voice doc.
Month 2 — Intake platform evaluation. If you don't already use Clio Grow, Lawmatics, or similar, evaluate two. Pilot one. Measure intake time before and after.
Month 3 — Document drafting pilot (if applicable). If your firm does real contract volume, pilot Spellbook, Harvey, or CoCounsel for a single attorney for thirty days. Measure drafting time, error rate, and attorney sentiment.
Month 4 — Research platform upgrade (if applicable). If you're currently doing research with consumer tools or basic Westlaw/Lexis, evaluate Lexis+ AI or Westlaw Precision with AI.
Month 6 — Knowledge management. Index past work. Set up search. This investment compounds for years.
What you should never automate
Legal advice. AI does not give legal advice. The attorney gives legal advice. AI helps the attorney prepare to give legal advice. The line matters.
Court appearances and signed filings. Always your work product, always your signature, always your verification.
Client communication where empathy matters. Bad news, difficult settlement discussions, sensitive personal matters — these are human conversations. AI can help you organize your thinking; it does not write the email or take the call.
Conflicts of interest analysis. Get human attention on this. AI assists with intake; it does not clear conflicts.
Ethical compliance generally. AI is not your ethics counsel.
Anything that exposes confidential client information to a tool without proper data handling agreements. No shortcuts here, no time-saving worth the cost.
Tools, by name
For a small firm starting from zero, here's a stack to evaluate:
- For non-confidential work: Claude (Anthropic) for nuanced writing, ChatGPT (OpenAI) for structured tasks. Use the paid tiers, not free, for better quality and clearer data handling.
- For confidential work: Harvey, Thomson Reuters CoCounsel, Lexis+ AI, Westlaw Precision with AI, Spellbook, or Clio Duo. Evaluate based on practice area and volume.
- For intake: Clio Grow, Lawmatics, or your existing case management system with AI features.
- For drafting: Spellbook (transactional), Harvey, CoCounsel, or your DMS's built-in features.
- For research: Lexis+ AI, Westlaw Precision with AI, vLex Vincent.
- For internal knowledge: Glean, NetDocuments with AI, iManage, or a well-organized SharePoint/Notion setup.
Monthly cost for a small firm running a real stack sits between $200 and $1,500 per attorney per month, depending on which specialty tools you bring in. The marketing-only side runs under $50 per month. The legal-specific tools usually pay back inside a quarter if the volume is there.
The short version
AI in a small law firm splits cleanly into work that touches client information (use legal-specific or enterprise tools, with verification on every output) and work that doesn't (use consumer tools freely, with voice anchoring).
The attorney's job is judgment, advocacy, and counsel. AI takes the drafting, formatting, intake, and research-support work off the desk so more of the attorney's day goes to the parts of the job that actually require an attorney.
If you'd like the system designed for your specific practice — your practice area, your jurisdiction, your existing tech stack — the Daring Brief is the place to start.