AI for small law firms: 7 jobs worth handing over in 2026

Small law firms get the most from AI on the work around the law, not the law itself: answering every new enquiry, booking consultations, preparing conflict checks, summarising documents, drafting routine letters for review, chasing clients for documents and keeping clients updated. Start with intake, because most firms lose more clients at the front door than they lose hours anywhere else. Use the AI already in your practice management software before buying more, keep client information out of personal AI accounts, and check every citation an AI produces. Below are the seven jobs worth handing over, what Clio and MyCase now do, the professional rules in the US and the UK, and a worked example for a four-lawyer firm. None of it is legal advice.

How are small law firms using AI in 2026?

Widely, and mostly at the desk. Clio's 2026 Legal Trends Report, published on 26 October 2026, found that 85% of legal professionals report some level of AI adoption at their firms, up from 79% a year earlier, and that the billable share of a lawyer's day is projected to reach 42%, up from 39%. The 8am 2026 Legal Industry Report, from more than 1,300 legal professionals, found 69% using general-purpose tools such as ChatGPT, mainly to draft correspondence, research, brainstorm and summarise documents, while fewer than half of firms train their people to use them responsibly.

What hasn't moved as fast is the front door. When Clio launched its intake agents in August 2026, it quoted its own research: 67% of firms don't respond to email enquiries, 48% don't respond to phone enquiries, and 79% of prospective clients expect a reply within 24 hours.

AI ON THE DESK, NOT AT THE DOOR

Law firms use AI. Many still don’t answer the enquiry.

• Legal professionals whose firm uses AI in some way, 202685%
• The same, a year earlier79%
• Use general-purpose AI tools such as ChatGPT69%
• Firms that don’t reply to an email enquiry67%

  The front door: where most small firms lose clients before any legal work starts.

• Firms that don’t respond to a phone enquiry48%

Show the data as a table

Measure	Share
Legal professionals whose firm uses AI in some way, 2026	85%
The same, a year earlier	79%
Use general-purpose AI tools such as ChatGPT	69%
Firms that don’t reply to an email enquiry	67%
Firms that don’t respond to a phone enquiry	48%

For a small firm, that gap is the opportunity. AI that helps a lawyer draft faster saves minutes. AI that makes sure the 22:00 enquiry gets a reply, a conflict check and a booked consultation wins matters.

7 jobs worth handing to AI in a small law firm

Three things sit outside the table on purpose. Legal research belongs to tools built for it, which check citations against real sources. Advice belongs to lawyers. And anything a court sees needs a lawyer who has read every line.

Intake first: what it should and shouldn't do

Intake is where AI earns its keep in a small firm, and where it most needs rules. A good intake assistant answers within a minute, at any hour, on the phone, the website and email. It says plainly that it isn't a lawyer and isn't giving advice. It collects facts, not opinions, checks the names it hears against your conflict list, offers real slots from your calendar, and leaves a summary on the lawyer's desk by morning.

TUESDAY · ONE ENQUIRY

AFTER HOURS

1. 22:10Website chat: “I was in a car accident last week. Do I have a case?”
2. 22:10“I can’t give legal advice, but I can take your details so a lawyer can look at this tomorrow.”
3. 22:12Asks for the date and place, injuries and treatment so far, the other driver’s name and insurer.
4. 22:14Checks the other driver’s name against the firm’s conflict list. No match.
5. 22:15Offers three consultation times from the lawyers’ calendars. The caller picks 09:30 Thursday.
6. 22:16Sends a confirmation and a secure link for photos and the police reference.

ON THE LAWYER’S DESK · 08:30

INTAKE SUMMARY

MATTER

Road traffic accident, 28 Oct

INJURIES

Neck and back, saw a GP twice

OTHER PARTY

Named, insurer known

CONFLICTS

None found. Lawyer to confirm

BOOKED

Thu 09:30, 30 min

NOT DONE

No advice given. No fee or outcome discussed

What it should never do: tell anyone they have a case, estimate what a claim is worth, promise an outcome, quote a fee you haven't set, or carry on a conversation about the merits. The summary card above lists what it didn't do for a reason. If your lawyers can see that at a glance, they can trust the rest.

Speed matters most where the client is choosing between firms tonight: personal injury, family, immigration, criminal defence and employment. If most of your work arrives by referral and the client already knows they want you, intake AI is a smaller win. If you're unsure, our guide to the cost of missed calls shows how to measure what slips through before you spend anything.

What Clio and MyCase already do

Check your practice management software first, because the leading products have added a lot in 2026.

• Clio now spreads AI across its products, as its own feature page describes them (checked November 2026). Clio Grow's Grow AI, launched in August 2026, answers phone, email and web chat, scores leads against the firm's own criteria and conflict rules, books consultations and follows up with leads that stall; its voice agent is US-only for now, and it charges $25 per lead it helps convert. Clio Work drafts and analyses, Vincent does legal research against curated sources and checks citations, and Clio's workflows and AI billing turn plain-language instructions and recorded time into tasks and draft bills. Several of these need particular plans or add-ons.
• MyCase brands its AI 8am IQ, with writing and document help and, on its higher plans, case and discovery assistants.
• Smokeball, PracticePanther, Filevine and CosmoLex in the US, and LEAP and Actionstep in the UK, have their own AI features too. Ask your account manager what your plan already includes before you pay for anything new.

The gap is usually between systems, not inside one. Enquiries arrive in a shared inbox, a web form and a phone that goes to voicemail; the conflict list lives in one place and the calendar in another; the client's documents land as email attachments. Connecting those, with the rules your firm already follows, is where most of the time goes, and it's the part we build.

Worked example: a four-lawyer personal-injury firm

To show how to weigh it up, here is a sample firm. Every number below is an assumption to replace with your own, not a result.

If two good matters a month slip away overnight, the question is simply whether an intake system costs less than two matters' fees. For most personal-injury, immigration and family firms the answer is obvious once the figures are written down. For a firm whose enquiries all arrive by referral in office hours, it may not be, and we would say so.

A typical first project for a firm like this takes a few weeks:

1. Map the front door. Every way enquiries arrive, who sees them, and how long replies take today, from your own logs.
2. Write the rules with you. The not-legal-advice wording, which practice areas to accept, the questions to ask for each, what triggers a hand-off to a person tonight, and what the assistant must never say.
3. Connect the systems. Intake to your practice management software, conflict list and calendars, so nothing is typed twice.
4. Run it alongside your team for two weeks, with a lawyer reading every summary.
5. Then go live after hours, and measure replies, bookings and retained matters against the month before.

The professional rules, in plain English

Both sides of the Atlantic now say much the same thing. This is a summary, not legal advice: read the guidance itself and your own regulator's rules.

In the US, ABA Formal Opinion 512 (29 July 2024) applies the existing Model Rules to generative AI:

• Competence (Rule 1.1): relying on AI output "without an appropriate degree of independent verification or review" can breach the duty of competence, and the lawyer stays fully responsible for the work.
• Confidentiality (Rule 1.6): a client's informed consent is needed before information relating to the representation goes into a self-learning AI tool, and boilerplate wording in an engagement letter is not enough.
• Communication (Rule 1.4): tell clients about your AI use when they ask, when their guidelines require it, when it affects your fee, or when its output will influence a significant decision.
• Supervision (Rules 5.1 and 5.3): managing lawyers must set clear AI policies and train lawyers and staff.
• Fees (Rule 1.5): bill hourly only for time actually spent, don't charge clients for learning a tool you'll use routinely, and a flat fee may become unreasonable if AI makes the work much quicker.

State bars have issued their own opinions too, so check yours.

In England and Wales, the SRA's warning notice on the misuse of AI (17 August 2026) says solicitors must make sure every case cited is genuine, relevant and verifiable; that putting client letters into a public tool such as ChatGPT places them in the public domain and risks privilege; that supervisors remain accountable for work done through others; and that failing to have proper regard to the notice risks disciplinary action. The SRA isn't against AI: in May 2025 it authorised Garfield.Law, an AI-driven firm for small debt claims, on conditions that included client approval of every step, named solicitors accountable for every output, and no AI-proposed case law. In Scotland, the Law Society of Scotland publishes its own guide for solicitors.

Put together, a small firm's AI policy needs only a page:

1. Which tools are approved, on business plans that don't train on your data, with a data processing agreement.
2. What never goes into an unapproved tool: client names, matter details, documents, and anything privileged.
3. When clients are told or asked, and the consent wording, which a lawyer at your firm should approve.
4. Who checks what: every citation, every figure and every document that leaves the firm.
5. Training, with a record of who has had it.

The citation problem won't go away on its own

General AI assistants predict likely-sounding text, and a case name with a citation looks like any other text. In Mata v. Avianca in 2023, two New York lawyers were sanctioned after filing a brief that cited six cases ChatGPT had invented. It didn't stop there. As of 5 November 2026, Damien Charlotin's AI Hallucination Cases database lists 2,266 court decisions worldwide involving AI-invented material, 898 of them involving lawyers and 1,535 in the US.

The fix is dull and reliable: use research tools that link every authority to its source, and have a person read every case before it's cited. The same goes for figures, dates and quotations in anything a client or court will read. Our plain-English guide to how RAG works explains why answers built from your own documents, with sources shown, are much safer than answers from a model's memory.

Will AI replace lawyers?

Not the parts clients pay for. Regulators on both sides of the Atlantic keep the lawyer accountable for anything AI produces, and clients with something at stake still want a person who knows the law to judge their situation and stand behind the answer. What AI replaces is the work around that: the unanswered enquiry, the retyped intake form, the first draft, the chase. Clio's 2026 report found lawyers recording more billable work as AI spreads, not less. The firms most at risk aren't the ones whose lawyers get replaced, but the ones whose phones go unanswered while a competitor's assistant books the consultation.

What we would do

For most small firms, in this order:

1. Measure the front door for two weeks: every enquiry, when it arrived and how long the reply took.
2. Switch on what you already pay for in Clio, MyCase or your own software, and write the one-page AI policy above.
3. Fix intake first, with not-legal-advice wording, conflict checks and booking, connected to your systems.
4. Then add the back office one job at a time: document summaries, a private assistant on your precedents, chasing, client updates.

That's how we work with firms: AI lead generation and AI integration for the front door, AI document processing and a private knowledge base behind it, and AI consultancy for the policy and training. We build the systems and the rules; the law stays with your lawyers. If you're choosing a general assistant for drafting, our comparison of ChatGPT, Claude, Gemini, Grok and Muse sets out which plans keep your data out of training. For how other professional firms are approaching it, see AI for accountants.