AI for Law Firms: What It Can (and Can't) Do for Solo and Small Firm Attorneys

Every attorney has been pitched an "AI-powered" tool in the last 12 months. Most are using AI in some form — even if it's just asking ChatGPT to draft a client email. But few have a clear picture of what these tools actually do well, what they do poorly, and where the professional liability starts.

This isn't a hype piece. AI is genuinely useful for attorneys in specific, well-defined tasks. It's also genuinely risky in others. The attorneys who get the most value from it understand both sides of that line.

What AI Can Actually Do for Your Law Firm Today

This isn't speculation about future capability. These are tools that exist, are being used by attorneys right now, and are producing real time savings.

The broad categories where AI adds value in a legal practice:

• Content generation — attorney bios, practice area pages, blog posts, marketing copy
• Document drafting — first drafts of standard documents, letters, and templates
• Legal research — case identification and summary (with significant caveats covered below)
• Administrative tasks — email drafting, meeting summaries, billing descriptions
• Client intake — AI-assisted lead qualification and routing

The key distinction that determines whether AI is helpful or dangerous: AI as a drafting assistant versus AI as a decision-maker. The first is where the value lives. The second is where the liability lives.

AI produces probabilistic output — the most statistically likely response given a prompt. For generating a first draft of an attorney bio, that's extremely useful. For making a legal judgment about whether to accept a settlement offer, it's inappropriate regardless of how confident the output sounds.

Content Generation — Websites, Bios, Marketing Copy

This is where AI excels for attorneys, and where the risk is lowest. Overcoming the blank page, generating structure, drafting options, maintaining consistent tone — these are AI strengths.

Attorney bios, practice area pages, blog posts, FAQs, and email templates are all strong AI use cases. The workflow is consistent:

1. Prompt with specific information about your practice and target client
2. Review the draft — what's accurate? What's too generic?
3. Edit heavily: add your voice, your specific experience, jurisdiction-specific details
4. Compliance review against your state bar's advertising rules
5. Publish

For the detailed step-by-step workflow on attorney bios and website content, see Using AI to Write Your Attorney Bio and Website Content.

Prompt Engineering for Attorneys — Getting Useful Output

The difference between AI producing usable drafts and producing generic filler is the quality of the prompt you give it. Most attorneys type a vague request, get a mediocre result, and conclude the tool is not useful. The tool is fine. The prompt needs work.

A prompt is the instruction you give the AI. The more specific and structured it is, the better the output. Here is the framework that consistently produces attorney-grade first drafts.

The ROCA Prompt Framework

Role — Tell the AI who it is. "You are a legal content writer for a solo family law attorney in Austin, Texas" gives the AI a persona and jurisdiction to anchor its output.

Objective — State exactly what you want. "Write a 300-word practice area page introduction for child custody services" is specific. "Write something about custody" is not.

Context — Provide background the AI does not have. Your years of experience, your target client demographic, your firm's positioning, the tone you want. The AI cannot guess these things — it will default to generic without them.

Audience — Who is reading this? "Parents going through divorce who are worried about losing time with their children" produces very different content than "attorneys evaluating my practice for referrals."

Example Prompts That Produce Usable Output

For an attorney bio:

You are a legal content writer. Write a 250-word professional bio for Sarah Chen, a solo family law attorney in Austin, TX with 12 years of experience. She focuses on high-asset divorce and child custody. Her clients are typically dual-income professionals. Tone: warm but authoritative — not salesy, not stuffy. Include a brief mention of her UT Law education and her approach to collaborative divorce. Write in third person.

For a practice area page introduction:

Write a 400-word introduction for a personal injury practice area page. The attorney is James Okoro, a solo practitioner in Denver, CO. His clients are people injured in car accidents, slip-and-falls, and workplace incidents. They are typically stressed, in pain, and unsure whether they have a viable claim. Tone: empathetic and direct. Focus on what the reader needs to know right now, not on selling the firm. Do not use phrases like "aggressive representation" or "fight for you."

For a client email response:

Draft a brief email response to a potential client who submitted a contact form about a custody modification. Acknowledge their inquiry, let them know the attorney will review their situation, and offer to schedule a 30-minute consultation. Tone: professional and reassuring. Keep it under 150 words. Do not provide any legal advice or case assessment.

Notice the pattern: every prompt includes role, objective, context, and audience constraints. The AI has guardrails. Without them, it fills the gaps with assumptions — and its assumptions about legal content are almost always too generic to use.

Iterative Refinement

The first output is rarely perfect. Use follow-up prompts to refine:

• "Make this more conversational — less corporate."
• "Remove any language that could be construed as a guarantee of outcomes."
• "Add a paragraph addressing clients who are concerned about cost."
• "Rewrite the opening to lead with the client's problem, not the attorney's credentials."

Each follow-up is cheaper and faster than rewriting from scratch. Three rounds of refinement typically produce something that needs only light editing before use.

Document Drafting — Contracts, Letters, Templates

AI is useful for generating first drafts of routine legal documents: engagement letters, simple agreements, demand letters, form motions, and document templates.

What works well: boilerplate language, structural frameworks, first drafts of standard documents in common practice areas. If you have a clear sense of what you want a document to accomplish, AI can get you to a workable first draft faster than starting from scratch.

What doesn't work well — and where you should not rely on AI without significant review: jurisdiction-specific nuance, complex litigation strategy, anything where getting it wrong has major consequences. AI doesn't know your jurisdiction's peculiarities, the judge's preferences, or the specific procedural posture of your case.

The right workflow for document drafting: AI produces a first draft from your template or prompt, you apply your legal judgment and state-specific knowledge, the document gets revised to match the actual facts and requirements of the matter, then it goes out under your signature and your professional responsibility.

Legal Research — The Promise and the Reality

This is the area with the highest potential and the highest documented risk. Be careful here.

The tools: Harvey, Casetext (now part of Thomson Reuters), Lexis+ AI, and Westlaw AI are the major players for legal-specific research AI. General-purpose tools like ChatGPT and Claude can also assist with legal research, though they lack real-time case law databases.

What they can do: Find potentially relevant cases faster than manual search, summarize holdings, identify legal issues and arguments, help structure an analysis of a legal question.

The hallucination problem: AI tools can and do confidently cite cases that do not exist. They fabricate case names, citation numbers, court names, and holdings. The output looks real. It reads authoritatively. The case is fictional.

The safe use of AI for legal research: Use it as a starting point to identify issues and candidate cases, then verify every citation independently in Westlaw or Lexis before it appears in any filing or client communication. AI research narrows your search — it does not replace your verification obligation.

Never cite a case in a legal filing without independently confirming that the case exists, says what you think it says, and is good law in your jurisdiction.

Client Intake — AI-Powered Routing and Follow-Up

AI is useful for initial website intake — qualifying leads at any hour, collecting basic matter information, and routing inquiries to the right practice area.

A chatbot on your website can answer common questions 24/7, collect initial lead information when you're not available, and flag urgent matters for priority follow-up. The AI doesn't replace your intake process — it extends the front end of it beyond business hours.

Automated follow-up sequences for inquiries that don't convert immediately are a legitimate use of AI-adjacent automation: if a potential client submits a contact form but doesn't schedule a consultation, automated follow-up emails (drafted with AI assistance) can re-engage them without manual effort.

The critical constraint: Any chatbot or automated intake system must be configured to avoid providing legal advice. It can gather information, answer factual questions about your practice, and route inquiries. It cannot tell a potential client what their legal options are, whether they have a viable case, or what they should do. Configure accordingly and review the outputs your system produces.

Administrative Work

This is where AI delivers the highest return for the least risk. Administrative tasks have lower stakes than legal work, and AI's drafting capabilities translate directly into time savings.

Email drafting. Describe the email you need to send, get a draft in seconds, edit in 30 seconds, send. For high-volume email work — client updates, follow-up communications, referral thank-yous — this compounds significantly over a month.

Meeting and consultation summaries. AI transcription tools (Otter.ai, Fireflies, and similar) can transcribe and summarize consultations and meetings, producing a record of what was discussed without 15–30 minutes of manual note-taking after every meeting. Get client consent to record before using any of these tools.

Billing descriptions. AI can help write clear, professional time entry descriptions from brief notes. Instead of "worked on discovery," you describe what you did and the AI drafts a billing-appropriate description. This alone reduces one of the most common time-tracking friction points.

Scheduling. Scheduling automation (Calendly, cal.com, and built-in calendar tools) isn't strictly AI in the technical sense, but it eliminates the scheduling back-and-forth that eats into administrative time for most practices.

Integrating AI Into Your Daily Workflow

The attorneys who get real value from AI are not the ones who use it occasionally for one-off tasks. They are the ones who have built it into repeatable workflows — specific points in their daily routine where AI handles the first draft or the grunt work, and the attorney handles the judgment.

The Morning Triage Workflow

Start each day by feeding AI your overnight intake submissions. Paste in the contact form details (with identifying information removed if using a non-confidential tool) and prompt: "Categorise this inquiry by practice area, urgency level (high/medium/low), and whether it falls within [your practice areas]. Flag anything that looks time-sensitive." This gives you a prioritised list in seconds instead of reading through each submission individually.

The Document Production Workflow

For any standard document you produce regularly — engagement letters, discovery requests, demand letters — build a reusable prompt template. The template should include your firm's standard language, the document structure you prefer, and placeholders for case-specific facts.

When a new matter requires the document, fill in the case-specific details and run the prompt. The AI produces a first draft that follows your template structure and incorporates the new facts. You review, revise, and finalise. Over time, you build a library of prompt templates that covers your most common documents.

This is not about speed alone. It is about consistency. Every engagement letter follows the same structure. Every demand letter hits the same points in the same order. AI enforces your template discipline in a way that starting from scratch each time does not.

The Client Communication Workflow

Client updates are one of the highest-volume, lowest-complexity tasks in most practices. After a hearing, deposition, or significant case event, dictate or type a brief summary of what happened and what comes next. Prompt the AI: "Draft a client update email based on these notes. Tone: professional, clear, reassuring. Do not include legal advice or case predictions. Keep it under 200 words."

Review the draft, add any case-specific nuance the AI missed, and send. What used to take 10-15 minutes of drafting now takes 2-3 minutes of review. Across 20 client matters, that is hours per week returned to billable or strategic work.

The Weekly Content Workflow

Block 30 minutes per week for content production. Use AI to draft one FAQ answer, one blog post outline, or one practice area page update. Over a month, that is four pieces of website content you would not otherwise have produced. Over a year, it is a substantial content library that drives organic search traffic.

The key is routine, not ambition. Attorneys who try to produce a 2,000-word blog post in one sitting burn out. Attorneys who produce 300-500 words per week in a repeatable AI-assisted workflow build a library.

What AI Shouldn't Do (Yet)

To be direct about where the line is:

• Provide final legal advice to clients. AI output is probabilistic, not authoritative. You are the attorney. The judgment is yours.
• Make settlement or strategy recommendations without attorney analysis. AI can summarize what settlement offers have looked like in similar cases. It cannot tell you whether to accept this one.
• Handle trust accounting. Trust accounting requires verified accuracy. AI produces probabilistic outputs. These are incompatible requirements.
• Draft complex litigation documents without heavy attorney revision. Motion practice in contested litigation is not a boilerplate exercise. AI drafts need substantial attorney input to reflect actual case strategy.
• Communicate directly with opposing counsel or the court. Every communication under your name is your professional responsibility.

Bar Ethics and AI

Most state bars are actively developing guidance on AI use. The ethical obligations that apply are the ones you already have — they just apply in a new context.

Duty of competence (Model Rule 1.1). The duty of competence now includes understanding the tools you use in your practice. You don't need to be an AI engineer, but you should understand what an AI tool does, what its failure modes are, and how to verify its output.

Confidentiality (Model Rule 1.6). When you submit client information to an AI tool, read the terms of service carefully. Does the tool use your inputs to train its model? Who can access the data? For sensitive client matters, use AI tools with explicit confidentiality agreements — enterprise tiers of major tools typically offer data processing agreements — or on-premise tools that don't transmit data externally.

Supervision (Model Rule 5.1 / 5.3). You cannot delegate your professional judgment to an AI tool. You are responsible for everything you submit to a court or client under your name, regardless of how it was drafted.

AI disclosure. Some jurisdictions are considering or have implemented requirements to disclose when AI was used to draft court filings. Check your jurisdiction's current rules and any pending guidance from your state bar.

Choosing the Right AI Tool for Your Practice

Not all AI tools serve the same purpose, and the right choice depends on what you are trying to accomplish.

General-purpose AI (ChatGPT, Claude, Gemini): These are your workhorse tools for content drafting, email composition, document templates, and administrative tasks. They are broadly capable, inexpensive or free at entry level, and require no legal-specific setup. ChatGPT is the most widely documented with the largest ecosystem of guides and prompt libraries. Claude tends to produce more nuanced writing and follows complex instructions more reliably — particularly useful for legal content where tone and precision matter. Start with one of these.

Legal-specific AI (Harvey, CoCounsel, Lexis+ AI, Westlaw AI-Assisted Research): These tools are trained on or connected to legal databases and designed for legal research, case analysis, and document review. They are more expensive, require more setup, and have steeper learning curves. Their advantage is access to real case law databases, which general-purpose tools lack. Their disadvantage is cost — most are priced for firms, not solo practitioners.

Transcription and meeting AI (Otter.ai, Fireflies.ai): Purpose-built for recording, transcribing, and summarising conversations. Useful for client consultations (with consent), depositions notes, and internal meetings. These are narrow tools that do one thing well.

The practical recommendation: Start with one general-purpose tool (ChatGPT or Claude) for content and administrative work. Add a transcription tool if you spend significant time in meetings or consultations. Evaluate legal-specific AI tools only after you are comfortable with general AI and have specific research workflows that would benefit from legal database integration.

Do not try to adopt everything at once. One tool, used consistently for a defined purpose, delivers more value than five tools used sporadically.

How to Start Without Overwhelming Yourself

The easiest path is to start with the lowest-risk, highest-value applications and add complexity from there.

Quick win 1: Use AI to draft or improve your attorney bio. This is low stakes (you review before publishing), clearly valuable (most attorney bios are weak), and a concrete starting point. The step-by-step workflow is in this post.

Quick win 2: Use AI to draft client email responses. Describe what you want to say, get a draft, edit for accuracy and tone, send. Do this for one week and track how much time it saves.

Quick win 3: Use AI to draft FAQ content for your website. Compile a list of the questions you're asked most often in consultations. Prompt AI to draft answers. Review for accuracy and bar compliance. Publish. This is a legitimate marketing and client service improvement.

Tool to start with: ChatGPT (widely documented, accessible) or Claude (often stronger on legal writing tasks, good at following structured prompts). Both are appropriate starting points.

What to avoid first: AI legal research tools. They have the highest learning curve and the highest consequence for errors. Get comfortable with lower-stakes AI use before trusting AI to assist with case law research.

AI doesn't replace attorney judgment. It reduces the time you spend on tasks that don't require it — so you have more time for the work that does.

Related reading: Using AI to Write Your Attorney Bio and Website Content | The Modern Law Firm Tech Stack