AI for Law Firms: Safe Ways to Automate Intake, Billing and Admin
AI for law firms, done safely: how small firms can automate intake, billing and admin within ABA ethics rules, plus real prices for 7 tools.
Most small law firms don't lose sleep over legal research.
They lose sleep over the phone that rings during a hearing, the intake form that sits in an inbox for two days, and the invoices nobody had time to send.
That's the work that quietly eats the week, and it's exactly where AI for law firms can help the most.
But a law firm isn't a coffee shop. You hold secrets, you hold client money, and your license depends on how both are handled.
So the real question isn't "Can AI do this?" It's "Can AI do this safely, inside the ethics rules?"
In this guide, we'll show you the safe ways a small firm can use AI and automation for client intake, billing and everyday admin, what the American Bar Association (ABA) and two state bars have said about it, and what real tools cost.
Two quick notes before we start.
First, this is a practical guide, not legal advice. Ethics rules differ by state, so always check your own state bar's rules and opinions before you change how your firm works.
Second, we checked every price on each vendor's official US pricing page on October 28, 2026, and where a vendor doesn't publish prices, we say so instead of guessing.
Here's what we'll cover:
What AI for law firms really means, in plain English
The ethics ground rules from the ABA and state bars
A simple traffic-light system for deciding what to automate
Step-by-step safe setups for intake, billing and admin
Real tools and prices, with worked cost examples
Questions to ask vendors, a one-page AI policy, and a 30-day rollout plan
AI for Law Firms: The Short Answer
AI for law firms means using software that can read, sort, summarize or draft, so your team spends less time on paperwork and more time on clients.
For a small firm, the safest wins are usually in the back office, not in legal judgment:
Intake: online forms, automatic replies, booking links and short summaries for the lawyer to review.
Billing: time-entry suggestions, invoice reminders and online payments that keep trust money separate.
Admin: first drafts of routine emails, document templates, meeting notes and status updates.
The ABA's main ethics opinion on generative AI, Formal Opinion 512, says lawyers can use these tools, but they must think about competence, confidentiality, client communication, supervision and fees (ABA Formal Opinion 512).
In plain terms: you can let AI help, but you stay responsible for everything it touches.
Our rule of thumb for the whole guide: AI can prepare, sort and draft, but a person at your firm checks, decides and sends.
What "AI for Law Firms" Really Means
The word "AI" gets stuck on everything these days, so let's split it into two kinds of tools you'll meet.
Plain automation vs generative AI
Plain automation follows fixed steps you set up in advance. Think of a vending machine: you press B4, and you get the same snack every time.
An example is a rule that sends a "We got your form" email the moment someone submits your intake form. It never improvises.
Generative AI (software that writes new text, like a summary or a draft, based on patterns it learned from huge amounts of writing) is different. Think of it as a very fast new intern who writes good first drafts but sometimes states things confidently that aren't true.
That last part matters. The State Bar of California warns that generative AI can produce plausible but inaccurate information, including citations to legal authority that doesn't exist, which are often called "hallucinations" (State Bar of California).
In one well-known federal case, a judge sanctioned two lawyers and their firm for filing fake citations created by generative AI, as the Florida Bar's ethics opinion recounts (The Florida Bar).
So here's the practical split:
Use plain automation for anything that must happen the same way every time, like reminders, receipts and routing.
Use generative AI for first drafts and summaries that a person will read before anyone relies on them.
Many of the safest wins in this guide are plain automation. That's a feature, not a weakness.
If you want a deeper look at why AI tools get things wrong, read our guide on why AI automation makes mistakes and how to prevent them.
Agentic AI: the new kind to watch
You'll also hear the phrase agentic AI, which means AI that can take actions on its own, like sending messages or moving files, not just write text.
California's 2026 guidance gives "autonomously facilitating client intake" as one example of what these systems can do (State Bar of California).
It says lawyers must not let these systems give legal advice or act for the lawyer without meaningful lawyer supervision and review (State Bar of California). Think of cruise control: handy, but your hands stay near the wheel.
Where small firms are today
If you haven't started yet, you're not alone.
In the ABA's 2024 Legal Technology Survey, 30.2% of attorneys said their offices were using AI-based tools. That figure was 24.1% for firms with 2 to 9 lawyers, and 17.7% for solo lawyers (ABA TechReport).
The top benefit lawyers expected was "saving time/increasing efficiency," picked by 54.4% (ABA TechReport).
The top worries were accuracy (74.7%), reliability (56.3%) and data privacy and security (47.2%) (ABA TechReport).
The Ethics Ground Rules Before You Automate Anything
Before we touch a single tool, let's set the ground rules.
The ABA Model Rules of Professional Conduct are a model that states adapt into their own rules, so the exact wording in your state may differ. The ABA says they "serve as models for the ethics rules of most jurisdictions" (ABA Model Rules). We cite the ABA versions here, plus guidance from Florida and California, because they're widely discussed.
Always check your own state's rules and ethics opinions, and when in doubt, call your state bar's ethics hotline if it has one.
Competence: know what your tools do
The ABA says a lawyer's duty of competence includes keeping up with "the benefits and risks associated with relevant technology" (ABA Model Rule 1.1, Comment 8).
Opinion 512 says lawyers don't need to become AI experts, but they do need a reasonable understanding of what a tool can and can't do (ABA Formal Opinion 512).
It also warns that relying on AI output without an appropriate degree of independent checking could violate the duty of competence (ABA Formal Opinion 512).
What this means day to day:
Try a tool on fake, made-up files before you use it on real ones.
Learn its known weak spots, like dates, names and numbers.
Decide who checks each type of output, and write it down.
Confidentiality: be careful what goes in
ABA Model Rule 1.6(c) says a lawyer must make reasonable efforts to prevent the unauthorized disclosure of, or access to, information relating to a client's representation (ABA Model Rule 1.6).
Opinion 512 goes further for self-learning AI tools, meaning tools that can use what you type to improve themselves, so your input might later shape answers for other users.
For those tools, the ABA says a client's informed consent, meaning agreement given after you've explained the risks in plain terms, is required before you put information relating to the representation into the tool. It adds that general boilerplate in an engagement letter (the written agreement that starts the lawyer-client relationship) is not enough (ABA Formal Opinion 512).
California's guidance says a lawyer must not put confidential client information into an AI tool that may present material risks to confidentiality or security, absent informed client consent. It also says reasonable efforts "require more than reliance on generalized marketing assurances" (State Bar of California).
Florida's opinion tells lawyers to research a program's policies on data retention, data sharing and self-learning (The Florida Bar).
Here's an everyday way to think about it. Typing client facts into a random free chatbot is like discussing a case loudly in a crowded elevator. You may never know who heard, but you can't take it back.
What this means day to day:
Read the tool's terms of use and privacy policy, and look for clear answers on training, retention and sharing.
Prefer business plans that state, in writing, that your data isn't used to train AI models.
Never paste client names, case facts or documents into a personal or free AI account.
Remember that duties also reach prospective and former clients. Opinion 512 points to Rules 1.18(b) and 1.9(c) on this (ABA Formal Opinion 512).
For a plain-English checklist on app permissions and data access, see our guide on whether it's safe to connect your business apps to AI automation tools.
Supervision: treat AI like a new assistant
ABA Model Rule 5.3 covers nonlawyer assistants. Its comments say lawyers must give assistants appropriate instruction and supervision, especially about not disclosing client information (ABA Rule 5.3 Comments).
The comments also name "using an Internet-based service to store client information" as an example of outside help a lawyer must make reasonable efforts to oversee (ABA Rule 5.3 Comments).
Opinion 512 says managing lawyers must establish clear policies on the firm's permitted use of AI, train their people, and do due diligence on vendors (ABA Formal Opinion 512).
Florida's opinion makes the comparison directly: lawyers who rely on generative AI for research, drafting, communication and client intake face many of the same risks as relying on an inexperienced or overconfident assistant (The Florida Bar).
So treat any AI tool like a new hire on their first week. You give clear instructions, you check their work, and you don't hand them the keys to the trust account.
Communication: telling clients what you use
ABA Model Rule 1.4 requires lawyers to keep clients reasonably informed and to explain matters enough for clients to make informed decisions (ABA Model Rule 1.4).
Opinion 512 says whether you must tell a client about AI use depends on the facts. You must disclose it if the client asks, and you should consult the client when AI use affects the fee or when its output will shape a significant decision. It suggests the engagement agreement is a logical place for disclosures (ABA Formal Opinion 512).
California's guidance also says lawyers must follow any client instructions that limit AI use (State Bar of California).
A good habit: write one short, plain paragraph in your engagement letter about how your firm uses technology, then talk about it in person when it matters. Just remember the ABA's point that boilerplate alone doesn't count as informed consent for self-learning tools.
Fees: bill honestly for AI-assisted work
ABA Model Rule 1.5 bans unreasonable fees and expenses, and says the basis of fees and expenses should be shared with the client, preferably in writing (ABA Model Rule 1.5).
Opinion 512 explains how that applies to AI (ABA Formal Opinion 512):
If you bill hourly, you bill the time you actually spent, including time writing instructions for the AI and reviewing what it produced.
AI built into ordinary software, like a grammar checker, is generally overhead, not a separate charge.
A third-party AI tool used for a specific client, priced per use, may be billed as an actual expense.
You may not charge a client for the time you spend learning a tool you'll use regularly.
Florida's opinion adds that if you can't work out the actual AI cost for a specific client, you shouldn't split a subscription across clients, and should treat it as overhead instead (The Florida Bar).
Advertising and chatbots
If your website uses an AI chatbot, advertising rules come into play too.
Florida's opinion says generative AI chatbots that talk with clients or others must include a disclaimer that the chatbot is an AI program, not a lawyer or employee of the firm (The Florida Bar).
It also warns against an "overly welcoming" chatbot that might give legal advice or accidentally create a lawyer-client relationship (The Florida Bar).
Even if you're not in Florida, those are sensible defaults. We'll come back to chatbots in the intake section.
A Simple Traffic-Light System for Law Firm Tasks
Rules are easier to follow when you can sort tasks quickly. We use a traffic light.
This is our own planning tool, not an official standard, so adjust it to your state's rules and your practice.
Green: automate now, with light checks.
Sending "we received your form" replies
Booking links and appointment reminders
Invoice reminders and payment receipts
Filling templates with details a person already checked
Turning your own notes into a tidy to-do list
Amber: AI drafts, a person reviews every time.
Summaries of intake forms for the lawyer
First drafts of routine client emails and status updates
Time-entry suggestions before they become bills
Meeting notes and call summaries
Pulling key dates out of documents
Red: keep it human.
Deciding whether to take a case
Giving legal advice or predicting outcomes
Anything sent to a court
Moving money in or out of a trust account
Final legal documents leaving the firm unchecked
Think of the amber lane like a newspaper's copy desk. Reporters file fast, but nothing goes to print until an editor reads it.
(This is a made-up example.) Harbor Point Estate Law is a solo practice with one part-time assistant. The lawyer prints the traffic light and tapes it next to the assistant's screen. When a new tool promises to "answer client questions automatically," the assistant checks the list, sees that legal advice is red, and the lawyer limits the tool to booking and reminders.
Safe Ways to Automate Client Intake
Client intake is everything that happens between "a stranger contacts your firm" and "a signed client with an open file."
It's usually a chain like this:
Inquiry → contact details → conflict check → consultation → decision → engagement letter → payment → open file.
Every arrow is a hand-off where a lead can slip away or a detail can get lost. That's why intake is often the first place firms automate.
We wrote a full guide on how to automate lead follow-up with AI, and our list of 20 small business processes you can automate with AI includes a short section on legal intake questionnaires.
Here, we'll focus on the law-firm-specific safety steps.
Step 1: Use an online intake form that asks less, not more
The instinct is to ask everything up front. For a law firm, that can backfire.
ABA Model Rule 1.18 covers prospective clients, meaning people who consult you about possibly hiring you. Even if you never take the case, you generally can't use or reveal what they told you (ABA Model Rule 1.18).
The same rule can disqualify you from representing someone else in the same or a substantially related matter if you received information that "could be significantly harmful" to the prospective client (ABA Model Rule 1.18).
One of the rule's exceptions depends on the lawyer having taken reasonable measures to avoid exposure to more disqualifying information than was reasonably necessary to decide whether to take the matter (ABA Model Rule 1.18).
In plain words: a long, detailed first form can load your firm with information you didn't need yet.
Think of the first form like a doctor's front desk asking for your name and the reason for your visit. The detailed history comes later, in the exam room.
A safer first form asks for:
Name and contact details
Type of matter, picked from a short list
Names of the other people or businesses involved, for the conflict check
Any urgent deadline, like a court date
How they'd like to be contacted
It also shows a short notice before they type anything. For example, that submitting the form doesn't make them a client, and that they shouldn't send confidential details yet.
Have your notice wording reviewed against your state's rules. We're showing the idea, not drafting it for you.
Most practice management tools, the main software where a firm keeps its cases, contacts, time and bills, include intake forms. MyCase lists Client Intake Forms on its Pro plan, and PracticePanther lists Intake Forms on its Business plan (MyCase) (PracticePanther).
Step 2: Run the conflict check before the deep dive
A conflict check is a search of your past and current clients and opposing parties to make sure taking the new matter wouldn't put you on both sides of a dispute.
Automation helps by moving the names from the form straight into your practice management system, so nobody retypes them.
The flow looks like this:
Form submitted → names copied into your system → assistant runs the conflict search → lawyer reviews any matches → only then book the full consultation.
Some tools build this in. MyCase lists Conflict Check Tracking on its Advanced plan, and Lawmatics lists advanced conflict checking on its Premium plan (MyCase) (Lawmatics).
Keep the final call human. A name match can be a false alarm, and a missed alias can be a real problem, so a person should look at every result.
Step 3: Send an instant reply and a booking link
Speed matters in intake. Someone who needs a lawyer often contacts more than one firm.
This is a green-light task. Plain automation can send a friendly reply within seconds:
"Thanks, we've received your request. A member of our team will review it. Here's a link to book a short call."
The booking link can show only the times your lawyer has set aside for consultations, and send reminders before the call.
Clio's pricing page says Clio Grow lets firms "book consultations and run intake with online forms and scheduling," and it's included in Clio's Elite plan or sold as an add-on (Clio).
Lawmatics lists client intake, custom forms, calendaring and e-signature on its Essential plan (Lawmatics).
Make sure the instant reply doesn't promise anything. "We've received your request" is safe. "We can definitely help you win" is not.
Step 4: Let AI summarize for the lawyer, not advise the prospect
This is where generative AI earns its keep, in the amber lane.
Once a prospect passes the conflict check and fills in a longer questionnaire, AI can turn three pages of answers into a short summary:
Who's involved
What happened, in date order
Key deadlines mentioned
Documents they say they have
Questions the lawyer may want to ask
The lawyer reads the summary before the call, then checks it against the original answers. The summary is a map, and the form is the territory.
What AI should not do at this stage is tell the prospect whether they have a case. Florida's opinion says a lawyer may not delegate to generative AI any act that could be the practice of law, and it applies its guidance on nonlawyer intake staff to chatbots: identify themselves as nonlawyers, stick to collecting facts, and refer legal questions to a lawyer (The Florida Bar).
Only use AI summaries inside a tool whose data terms you've checked, as we covered in the confidentiality section.
Step 5: Send the engagement letter and collect the retainer
After the lawyer decides to take the matter, plain automation can handle the paperwork:
Lawyer approves → engagement letter fills from a template → client signs online → payment link sent → file opens automatically.
Rule 1.5 says the scope of the representation and the basis of your fee should be communicated, preferably in writing, and contingent fee agreements must be in a writing signed by the client (ABA Model Rule 1.5).
A template keeps that language consistent every time. The lawyer still reviews each letter before it goes out, because scope varies from case to case.
MyCase Pro lists unlimited e-signature and document generation, and PracticePanther Business lists native e-signature with unlimited sends (MyCase) (PracticePanther).
Website chatbots and AI receptionists: extra care
AI receptionists and website chatbots can answer the phone or chat at 9 p.m. when your office is closed.
We compare options in our guides to the best AI receptionist for small business and AI receptionist vs human receptionist.
For a law firm, add these guardrails, many of which come from Florida's opinion (The Florida Bar):
The bot says up front that it's an AI program, not a lawyer or staff member.
It collects facts and contact details only.
It never says whether someone has a case or what they should do.
It asks early whether the person already has a lawyer. Florida suggests screening questions that limit the chatbot's conversation if someone is already represented (The Florida Bar).
It shows clear notices that limit what the conversation means, so nobody believes they've hired you by chatting.
It hands anything urgent, like an arrest or a court date tomorrow, to a person.
California's guidance adds another point: AI used to screen clients may carry a risk of unlawful or unethical bias (State Bar of California). So don't let a bot quietly reject people. Let it collect and route, and let a person decide.
Worked intake example: a three-person family law firm
(This is a made-up example.) Birchwood Family Law has two lawyers and an office manager. They get about 30 new inquiries a month.
Before automation, each inquiry took about 20 minutes of the office manager's time: returning calls, emailing forms, chasing missing answers, and typing details into the system.
That's 30 × 20 = 600 minutes, or 10 hours a month.
Here's their new flow:
Website form (short) → instant reply with booking link → names go into the system → office manager runs the conflict check → longer questionnaire for cleared prospects → AI summary for the lawyer → lawyer calls.
If that cuts the office manager's time to about 8 minutes per inquiry, they save 30 × 12 = 360 minutes, or 6 hours a month.
These are planning numbers for the example, not results from a study. Time yours for two weeks before and after so you know your real savings.
The lawyers still decide who to take on, and nothing goes to a prospect except the reply template and booking link they approved.
Safe Ways to Automate Billing
Billing is where small firms quietly lose money. Time goes unrecorded, invoices go out late, and payments trickle in.
It's also where the rules are strictest, because Rule 1.15 governs client money (ABA Model Rule 1.15).
Capture time before it disappears
Every lawyer knows the feeling of reaching Friday and trying to remember what Tuesday afternoon was spent on.
Some tools track activity for you. Smokeball describes AutoTime as its "automatic time and activity tracking feature," included on its Prosper+ plan and available as an add-on for Grow users (Smokeball).
Think of activity tracking like a car's trip odometer. It records the miles as you drive, so you don't have to guess at the end of the day.
But a record of activity isn't a bill. A tool might log 40 minutes with a document open while you were really on a call about another matter.
So the rule is simple: tracked time is a suggestion, and a lawyer approves every entry before it reaches an invoice.
Let AI tidy time entries, then review them
Generative AI can turn rough notes like "call w/ client re: custody sched" into a clearer entry like "Telephone conference with client regarding custody schedule."
That helps clients understand their bills, which supports the communication duties in Rule 1.4 (ABA Model Rule 1.4).
But check two things every time:
The time matches what really happened. Opinion 512 says hourly billing must reflect the time actually spent (ABA Formal Opinion 512).
The description doesn't make the work sound bigger than it was. A polished sentence must still be a true one.
Florida's opinion says AI efficiency must not lead to falsely inflated claims of time, and suggests firms may consider flat or contingent fees so both lawyer and client benefit from the efficiency (The Florida Bar).
Automate invoices and reminders
This is green-light work for plain automation.
A typical setup:
Approved time → invoice built on the 1st of the month → emailed with a payment link → reminder at 7 days → second reminder at 14 days → staff call at 21 days.
The reminders go out on schedule, politely, every time. Nobody has to feel awkward about chasing a client.
Online payments and trust accounts
This part needs the most care.
ABA Model Rule 1.15 says lawyers must keep client property separate from their own. Advance fees and expenses go into a client trust account and can be withdrawn only as fees are earned or expenses incurred (ABA Model Rule 1.15).
A trust account is a bank account that holds money belonging to clients, not to the firm. Think of it like a coat check. You're holding the coat, but it isn't yours to wear.
Many small firms use an IOLTA (Interest on Lawyers' Trust Accounts), a pooled trust account for small or short-term client funds, where the bank sends the interest to a state program that mainly funds civil legal aid (ABA IOLTA).
Why does this matter for automation? Because a general payment tool might not know the difference between a retainer (money paid up front before the work is done), which belongs in trust, and a payment for an invoice you've already earned, which can go to your operating account.
That's why law firms usually choose payment tools built for lawyers. LawPay, for example, lists "Trust account protection and IOLTA compliance" among its features (LawPay).
MyCase lists IOLTA trust account protection with its LawPay Payments, and PracticePanther lists three-way trust reconciliation (MyCase) (PracticePanther).
A three-way reconciliation means checking that three records agree: your bank statement, your trust account ledger, and the total of each client's individual balance. It's like counting the cash drawer, the receipt roll and each order slip, and making sure all three match.
Rule 1.15 also requires complete records of trust funds to be kept, for five years after the representation ends in the ABA's model version (the number appears in brackets because states set their own period) (ABA Model Rule 1.15).
Our red-light rule stands: software can record, match and flag trust transactions, but a person approves every movement of trust money. Check your state's trust account rules with your bar before you change anything.
If you also want AI help with your general books, our guide to AI accounting software explains what to look for.
Charging for AI tools
If your firm pays for an AI tool, can you pass the cost on to clients? We covered the ABA's view in the fees section above. Two state bars add detail.
California's guidance says general AI subscriptions typically count as overhead. It allows matter-specific AI costs to be charged if the fee agreement clearly discloses it, the charge reflects actual cost, and there's no markup without the client's informed written consent (State Bar of California).
Florida's opinion says a lawyer should tell a client, preferably in writing, if the lawyer intends to charge the actual cost of using generative AI (The Florida Bar).
The simplest path for most small firms: treat AI subscriptions as overhead, and put any exceptions in writing in the fee agreement.
Safe Ways to Automate Admin
Admin is the "everything else" pile: email, documents, notes, data entry and keeping clients updated.
None of it is glamorous, but it's where many small firms find their easiest time savings.
Email: drafts, not sending
A lot of law firm email is routine: confirming appointments, asking for documents, sharing next steps.
An AI email assistant can sort your inbox and write first drafts of replies for you to edit. We explain how these tools work, and their risks, in our guide to the AI email assistant.
For a law firm, add three rules:
AI drafts, but a person reads and sends every client email.
Use an email tool on a business plan whose terms you've checked, not a personal account.
Double-check the recipient. Autocomplete sending a draft to the wrong "John" is a confidentiality problem, AI or not.
Rule 1.6(c) asks for reasonable efforts to prevent unauthorized disclosure, and a misdirected email is one of the most common ways that happens (ABA Model Rule 1.6).
Document automation: templates that fill themselves
Document automation means building a template once, with blanks for names, dates and details, and letting software fill those blanks from a form or your case file.
It's like a mail merge for wedding invitations. You design the card once, and each guest's name drops into place.
This is mostly plain automation, not generative AI, which makes it predictable. The same answers always produce the same document.
Gavel is one tool built for this. Its pricing page says it helps firms "turn intake into documents," with Word and PDF automation and a Clio Manage connection on its Lite plan (Gavel).
Good candidates include engagement letters, simple wills intake packets, demand letter shells, fee agreements and standard notices. The lawyer still reviews the finished document, because a template can't spot a fact pattern that doesn't fit.
Meeting notes and call summaries
AI note-takers can record a meeting and produce a summary with action items. We cover how they work in our guide to AI meeting notes.
In a law firm, slow down here, because recording a client conversation raises extra questions:
Recording consent laws differ by state, so check yours before you record anyone.
Recording and summarizing a client meeting means sending sensitive conversations to the tool's company, so confidentiality rules apply.
A safe starting point is using AI to tidy notes you typed yourself after a meeting, inside a tool whose data terms you've checked.
Data entry: stop retyping the same details
Small firms often type the same client details into four places: the intake form, the case file, the billing system and the calendar.
Connecting them, or using one system for most jobs, removes the retyping. See our guide on how to automate data entry with AI for the general approach.
For legal work, one extra check matters: dates. A wrong date in a calendar can mean a missed deadline. Treat any automatically created deadline as a draft until a person checks it against the court order or rule.
Client status updates
Rule 1.4 says lawyers must keep clients reasonably informed about the status of their matter and promptly comply with reasonable requests for information (ABA Model Rule 1.4).
"Any update on my case?" emails eat a lot of time, and they usually come from clients who feel in the dark.
Automation can help in two ways:
A client portal, where clients see documents, invoices and next steps themselves. MyCase and PracticePanther both list client portals (MyCase) (PracticePanther).
Draft updates at milestones, like "We've filed your petition," prepared by AI or a template and sent after a person checks them.
AI for Law Firms: Tools and Prices Compared
Here are seven tools a small firm might consider. They're not the only options.
We checked prices on each vendor's official US pricing page on October 28, 2026. Prices are in US dollars, per user per month unless we say otherwise, and they currently apply but can change.
All-in-one practice management
These tools are the "home base" for your cases, contacts, time, billing and intake.
Clio (best for firms that want a large, well-known platform): Plans: Starter, Core, Signature and Elite. Cost: Clio's FAQ says plans start at $49 per user per month, and Starter shows "Starting at $49/user." Core, Signature and Elite show "Get pricing," so those are by quote. The page doesn't say whether $49 is for monthly or yearly billing, so ask. Clio Grow, its intake product, is included in Elite and sold as an add-on for Core and Signature, with pricing from sales. Why: AI is included from Core up, there's a 7-day free trial with no credit card, and its FAQ says your firm's data is never used for AI training (Clio).
MyCase (best for firms that want clear public prices): Plans: Basic $60 monthly or $50 per month paid yearly; Pro $120 monthly or $100 paid yearly; Advanced $150 monthly or $130 paid yearly. Cost: MyCase Accounting is $39 per user per month on Pro and Advanced, and LawPay Payments has no monthly fee. Why: Pro adds intake forms, automated workflows, e-signature and an AI writing and document assistant; Advanced adds an AI case assistant and conflict check tracking. There's a 10-day free trial with no credit card (MyCase).
PracticePanther (best for firms that want texting and intake forms at a mid-range price): Plans: Solo $59 monthly or $49 billed annually; Essential $79 or $69; Business $99 or $89; Business Pro $124 or $114. Cost: the page says annual billing saves $120 per user per year. Why: Business adds intake forms, two-way texting and unlimited e-signature; Business Pro adds full trust and operating accounting (PracticePanther).
Smokeball (best for firms that want automatic time tracking): Plans: Bill, Boost, Grow and Prosper+. Cost: by quote. Smokeball says pricing depends on users, plan, contract term and optional products. Why: AutoTime automatic time tracking is included on Prosper+, along with intake and workflows (Smokeball).
Intake and growth
Lawmatics (best for firms focused on intake and marketing): Plans: Essential, Premium and Enterprise. Cost: by quote. Lawmatics says it offers "tailor-made pricing plans" and asks firms to request a custom quote. Why: Essential includes client intake, custom forms, calendaring and e-signature; Premium adds marketing automation, text messaging and advanced conflict checking (Lawmatics).
Documents
Gavel (best for turning intake answers into documents): Plans: Lite $99 per month billed monthly or $83 per month billed annually; Standard $250 or $210; Pro $350 or $290; Scale from $417 per month, billed annually, through sales. Cost: these are per account, with seat limits. Lite includes 1 builder seat, Pro includes 2 builder seats and 3 organization user seats, and extra seats on Pro and Scale cost $50 per builder or $35 per organization user per month. Why: Word and PDF automation, a Clio Manage connection on every plan, and a 7-day free trial (Gavel).
Payments
LawPay (best for firms that need card and bank payments with trust protection): Plan: $19 per month, no contract, billed monthly. Cost: card payments are 2.99% plus $0.30 per transaction for Visa, Mastercard and Discover; American Express is 3.90% plus $0.30; eCheck, a payment pulled straight from the client's bank account, is 1%. LawPay also lists a pass-through fee allocation of $7.99 per month, an extra 1.75% for foreign cards, and custom pricing on request. Why: trust account protection and IOLTA compliance, plus unlimited users (LawPay).
What about general AI chatbots?
General AI assistants can help with non-confidential work, like outlining a blog post or rewording a generic checklist.
Opinion 512 notes that using AI only to generate ideas, without client information, doesn't require client consent (ABA Formal Opinion 512).
If you want to go further, use a business plan with clear data terms. For example, OpenAI's ChatGPT Business plan lists "No training on your business data by default" and is priced per user per month, with monthly and annual options (OpenAI). The price didn't display when we checked, so confirm it on that page.
Worked Examples: What AI for Law Firms Really Costs
Let's put real prices into real-world setups. The firms are made up, but the prices are the ones we checked on October 28, 2026.
Example 1: A solo lawyer on MyCase Pro
(This is a made-up example.) Harbor Point Estate Law is one lawyer who handles most admin alone, so the firm buys one user seat.
Paying monthly: $120 × 12 = $1,440 a year.
Paying yearly: $100 × 12 = $1,200 a year.
Saving with yearly billing: $240, which matches the saving MyCase shows (MyCase).
If you later add an assistant, give them their own seat rather than sharing a login, so you can always see who did what.
Example 2: A three-person firm on PracticePanther Business
(This is a made-up example.) Birchwood Family Law needs 3 users: two lawyers and the office manager.
Billed monthly: 3 × $99 × 12 = $3,564 a year.
Billed annually: 3 × $89 × 12 = $3,204 a year.
Saving: $360 a year, which is the $120 per user PracticePanther mentions, times three (PracticePanther).
Remember the intake example earlier: if automation really saves the office manager about 6 hours a month, that's about 72 hours a year of admin time back. Your own numbers will differ, so measure before and after.
Example 3: Adding document automation with Gavel
(This is a made-up example.) A small estate planning firm wants questionnaires that turn into draft documents.
Gavel Standard billed monthly: $250 × 12 = $3,000 a year.
Gavel Standard billed annually: $210 × 12 = $2,520 a year.
Saving: $480 a year (Gavel).
Gavel describes annual plans as "2 months free" (Gavel). On our math, the Standard plan saves $480, which is a little less than two months at the monthly rate ($500), so check the exact figure for the plan you pick.
Example 4: The real cost of card payments
(This is a made-up example.) A client pays a $2,500 retainer through LawPay.
By Visa or Mastercard: $2,500 × 2.99% + $0.30 = $75.05.
By American Express: $2,500 × 3.90% + $0.30 = $97.80.
By eCheck: $2,500 × 1% = $25.00.
Plus the $19 monthly fee, which is $228 a year (LawPay).
Who pays the processing fee, and from which account it's taken, is a trust accounting question. Ask your payment provider how fees are handled for trust deposits, and check your state's trust account rules.
12 Questions to Ask Before You Choose a Tool
Bring this list to every demo. Ask for answers in writing where you can.
Is our data used to train your AI models, or anyone else's?
How long do you keep our data, and can we delete it?
What happens to our data if we cancel? Can we export it, and for how long?
Which outside AI companies do you send our data to, and under what agreements?
Will you notify us if there's a security breach?
Do you have an independent security report, like SOC 2 (an outside audit of a company's security controls)?
Can we turn AI features off for certain users or matters?
Does every AI action leave a record of who did what and when?
Can the AI send anything to clients or courts without a person approving it?
How do you handle trust account payments and fees?
Is the price billed monthly or yearly, and what does each user seat include?
Can we test it with made-up data before we put real client files in?
These questions mirror what the guidance asks for. Florida's opinion, borrowing from its cloud computing guidance, recommends checking a provider's confidentiality obligations and breach notice, its reputation and security, and whether it keeps information after you stop using the service (The Florida Bar).
Write a One-Page AI Policy for Your Firm
Opinion 512 says managing lawyers must establish clear policies on the firm's permitted use of AI (ABA Formal Opinion 512). California's guidance also calls for clear policies and periodic training (State Bar of California).
A small firm doesn't need a 30-page manual. One page that everyone actually reads is better.
Here's a simple outline to adapt with your own judgment and your state's rules:
Approved tools: list each tool by name and what it's approved for.
Never allowed: client information in personal or free AI accounts.
Review rule: every AI draft is checked by a named person before it's used or sent.
Red-light list: legal advice, court filings, trust money and case decisions stay human.
Client communication: how and when we tell clients about AI use, and how we follow any client's request to limit it.
Billing: AI subscriptions are overhead; any AI cost passed to a client is agreed in writing first; we bill only actual time.
Incidents: who to tell if something goes wrong, like a misdirected email or a wrong AI summary.
Review date: we revisit this policy every six months.
Think of it like the safety card in an airplane seat pocket. It's short, it's plain, and it's there before anything goes wrong.
A 30-Day Rollout Plan for a Small Firm
Don't switch everything on at once. Roll out in small steps so you can catch problems early.
Week 1: Measure and choose
Time your current intake, billing and admin tasks for one week. Pick one task to improve first, usually intake replies or invoice reminders.
Our small business automation roadmap explains how to choose a first project by time saved and risk.
Week 2: Test with made-up data
Start free trials and run fake clients through the whole flow. Check what each tool stores, what it sends, and who can see it.
Read the terms of use and privacy policy for each tool, and write down what you found.
Week 3: Go live with green-light tasks
Turn on instant replies, booking links, reminders and invoice schedules. Keep a person checking everything for two weeks.
Week 4: Add one amber-light task
Add AI intake summaries or time-entry drafts. Compare the AI version against the original every time for the first month.
At the end of the month, measure again and decide what to keep, change or drop.
(This is a made-up example.) Birchwood Family Law follows this plan. In week 2, a test reveals that their reminder emails show the client's matter name in the subject line. They change it to a neutral subject, so a client's shared family inbox doesn't reveal the case, before a single real email goes out.
Common Mistakes to Avoid
Watch out for these:
Using a personal AI account for client work. It's the fastest way to lose control of confidential information.
Letting a chatbot sound like a lawyer. Make sure it says it's AI and sticks to collecting facts.
Asking for the whole story on the first form. Collect conflict-check details first.
Trusting automatic time entries. They're suggestions until a lawyer approves them.
Mixing trust and operating money in a general payment tool. Use tools built for law firm trust accounting, and keep a person approving trust transactions.
Skipping the engagement letter update. If AI use affects fees or a client's decisions, talk about it.
Frequently Asked Questions About AI for Law Firms
What is AI for law firms?
It's software that can read, sort, summarize or draft text to help with legal and business tasks. For small firms, the safest uses are usually intake, billing and admin, with a person reviewing anything important.
Is it ethical for lawyers to use AI?
The ABA's Formal Opinion 512 says lawyers may use generative AI if they meet their duties of competence, confidentiality, communication, supervision and reasonable fees (ABA Formal Opinion 512). Your state's rules and opinions may add more, so check them.
Do I need client consent to use AI?
It depends on the tool and the facts. The ABA says informed consent is required before putting information relating to a representation into a self-learning AI tool, and that boilerplate in an engagement letter isn't enough (ABA Formal Opinion 512). Using AI only for ideas, with no client information, doesn't require consent under the same opinion.
Can AI do client intake for a law firm?
AI and automation can collect contact details, send replies, book consultations and summarize answers for the lawyer. They shouldn't give legal advice or decide whether to take a case. Florida's opinion says intake chatbots should identify themselves as AI and avoid legal advice (The Florida Bar).
Can a law firm bill clients for AI?
You can bill the time you actually spend, including reviewing AI output, if you bill hourly. General AI subscriptions are usually treated as overhead, while matter-specific per-use costs may be billable as expenses if handled properly (ABA Formal Opinion 512). Put any AI charges in writing in the fee agreement.
What is the best AI software for small law firms?
There's no single best tool. MyCase has clear public prices, Clio is a large platform with AI from its Core plan up, PracticePanther is strong on texting and intake forms, and Smokeball focuses on automatic time tracking. Pick based on your biggest time drain and test with made-up data first.
How much does AI for law firms cost?
On October 28, 2026, practice management plans we checked started at $49 per user per month for Clio Starter, and $50 per user per month paid yearly for MyCase Basic (Clio) (MyCase). Some tools, like Smokeball and Lawmatics, are by quote. Prices can change, so check the vendor's page.
Does AI make mistakes with legal information?
Yes. Opinion 512 cites a study that found legal research AI tools hallucinated between 17% and 33% of the time (ABA Formal Opinion 512). Always verify facts, dates, names and citations.
Final Verdict
AI for law firms works best when it takes the paperwork off your plate and leaves the judgment with you.
For most small firms, the safest order is:
Start with plain automation: instant intake replies, booking links, invoice reminders and online payments built for trust accounts.
Add AI drafts next: intake summaries, time-entry wording, routine emails and status updates, always reviewed by a person.
Keep the red lane human: legal advice, case decisions, court filings and trust money.
Pick tools that state in writing how they handle your data, test them with made-up files, and write a one-page policy before you go live.
And because every state is different, check your own state bar's rules and opinions before you change how your firm works.
Done carefully, AI can give a small firm back hours every month, without putting your clients' secrets or your license at risk.
To see more tasks you can hand off safely, read our guide to 20 small business processes you can automate with AI.
References
https://www.calbar.ca.gov/Portals/0/documents/ethics/Generative-AI-Practical-Guidance.pdf
https://www.floridabar.org/etopinions/opinion-24-1/
https://vividalgo.com/why-ai-automation-makes-mistakes-and-how-to-prevent-them
https://vividalgo.com/is-it-safe-to-connect-your-business-apps-to-ai-automation-tools
https://vividalgo.com/how-to-automate-lead-follow-up-with-ai
https://vividalgo.com/20-small-business-processes-you-can-automate-with-ai
https://www.mycase.com/pricing/
https://www.practicepanther.com/pricing/
https://www.lawmatics.com/pricing
https://vividalgo.com/best-ai-receptionist-for-small-business
https://vividalgo.com/ai-receptionist-vs-human-receptionist
https://www.smokeball.com/pricing
https://www.americanbar.org/groups/interest_lawyers_trust_accounts/overview/
https://www.lawpay.com/pricing/
https://vividalgo.com/ai-accounting-software
https://vividalgo.com/ai-email-assistant
https://vividalgo.com/ai-meeting-notes-how-small-businesses-can-stop-taking-notes-by-hand
https://vividalgo.com/how-to-automate-data-entry-with-ai
https://vividalgo.com/small-business-automation-roadmap-what-to-automate-first-and-what-it-costs