AI Agents for Lawyers: What They Can Do and What Must Stay Human

AI agents for lawyers have moved quickly from demos to products that firms can buy. During 2026, Thomson Reuters, Clio and others began selling tools that plan and complete several steps of legal work, rather than just answering a prompt.

At the same time, courts and regulators have become blunter about the risks. In June 2025, the High Court of England and Wales warned lawyers after fake case citations appeared in court documents. Professional bodies have also published guidance on confidentiality and checking AI output.

This guide covers what these agents can realistically do, which documented tools exist, what must stay with a qualified lawyer, and how a small firm can start carefully. It is general information, not legal or ethics advice. So check your own jurisdiction’s rules and your regulator’s guidance. TheJusGrow has not tested these products for this article.

What are AI agents for lawyers?

An AI agent is software that can break a goal into steps and work through them, often using other tools along the way. For instance, instead of asking a chatbot to summarise one contract, you might ask an agent to review fifty contracts, flag unusual clauses and put the results in a table.

In legal practice, that usually means one of three things. First, an agent built into a legal research platform. Second, an agent built into practice management software. Third, a general assistant such as ChatGPT, Claude or Microsoft Copilot, used inside a firm’s own accounts.

For AI agents for lawyers, the difference matters because the data, the sources and the contract terms differ for each. For a broader primer, our guide to agentic workflows explains how multi-step AI tasks work in everyday terms.

What can AI agents for lawyers do today?

Based on vendors’ published material, today’s tools cluster around five jobs. The table below describes what the companies say their products do, so treat it as a list of claims rather than a performance test.

Task	What the agent does (company descriptions)	Example product
Research and briefs	Moves from research to a first-draft brief, proposing facts, arguments and authority	CoCounsel Legal’s Westlaw Brief Builder
Bulk document review	Reviews up to 10,000 documents against up to 100 questions and returns a table	CoCounsel Legal’s Tabular Analysis
Citation checking	Checks whether cited authority supports a stated legal point	CoCounsel Legal’s Deep Research Verify
Matter work	Handles goal-style requests about a matter while showing its progress	Clio Work
Client intake	Answers enquiries by phone, email and chat, scores leads and books consultations	Clio Grow AI

Many AI agents for lawyers reach the United States first. For example, Thomson Reuters said in August 2026 that its next-generation CoCounsel Legal was available in the US, while Canada, the UK and Australia were expected later in 2026.

Which AI agent tools exist for law firms?

The market is moving fast, so the products below are examples drawn from their makers’ announcements and from legal technology reporting. They are not recommendations, and we have not ranked them.

Legal research platforms

Thomson Reuters launched the next generation of CoCounsel Legal on August 20, 2026. It describes the product as combining reasoning, planning and execution across multi-step legal workflows. Its listed features also include a drafting agent for Microsoft Word that uses Practical Law content, plus matter-specific workspaces that combine a firm’s own documents and precedents. However, the launch announcement did not include pricing.

Practice management software

Clio added agentic capabilities to Clio Work in April 2026, according to legal technology journalist Bob Ambrogi at LawSites. He reported that lawyers can give goal-style instructions, watch the tool’s progress in real time, and interrupt or redirect it mid-task.

Then, in August 2026, Clio launched Grow AI intake agents. LawSites reported that firms pay $25 per converted lead, that the voice agent works only in the US for now, and that firms can set a monthly spending target to keep costs predictable.

General assistants

Many smaller firms start with a general assistant instead. Microsoft, for example, said in September 2026 that Skills in Microsoft 365 bring specialised expertise into its apps, including legal skills in Word. Such tools can help with plain-language letters, summaries and checklists. Still, they are not legal research databases, so every citation needs checking against an authoritative source.

Some firms are also watching new standards that let agents connect to legal data in a structured way. Our explainer on the Legal Context Protocol covers one of these.

What must stay human when lawyers use AI agents?

Professional rules do not disappear because a tool is clever. In July 2024, the American Bar Association published Formal Opinion 512, its first formal ethics guidance on generative AI. It ties AI use to existing duties, including:

• Competence: understanding the benefits and risks of the tools you use.
• Confidentiality: protecting client information, which can mean getting a client’s informed consent before entering it into certain AI tools.
• Communication: consulting clients about how their matter will be handled, which can include AI use.
• Fees: the ABA says lawyers may bill for time spent entering data and reviewing AI output, but generally not for learning a tool.

Similarly, the Law Society’s guide for solicitors in England and Wales says professional obligations apply whether a solicitor uses AI personally or supervises someone who does. It also advises against putting confidential information into generative AI tools without direct control and oversight.

The courts have made the stakes clear. In the Ayinde and Al-Haroun judgment of June 2025, Dame Victoria Sharp, President of the King’s Bench Division, dealt with filings that cited fabricated authorities. In one case, 18 of 45 cited authorities did not exist. As a result, the court stressed that lawyers have a continuing duty to check accuracy against authoritative sources.

So in practice, a sensible firm keeps these decisions with a person:

• final legal judgment and the advice a client relies on;
• anything filed with a court or sent to the other side;
• decisions on conflicts of interest and whether to take a matter;
• agreements on fees and engagement terms;
• choices about which client data goes into which tool.

How can a small firm start with AI agents?

With AI agents for lawyers, a cautious pilot beats a firm-wide rollout. So here is a starter workflow that follows the guidance above:

1. Pick a low-risk task. Internal knowledge summaries or first drafts of standard letters are safer than court filings.
2. Read the contract. Check the vendor’s terms on data storage, model training and confidentiality, and note which security statements are only company claims.
3. Test with fictional data. The Law Society suggests fictional data for testing, rather than personal data or client information.
4. Set review rules. Decide who checks each output and how citations get verified.
5. Tell clients when needed. Update engagement letters or policies if your rules require it.
6. Log tool use. Record which tool worked on which matter and who reviewed the output.
7. Review after a month. Keep, change or drop the tool based on errors found and time saved.

A worked example shows how this can look. Imagine a two-lawyer employment practice that wants faster intake. First, it sets up an intake agent to collect contact details and a short description of the problem outside office hours. Then a lawyer reviews each enquiry the next morning, runs the firm’s usual conflict check and decides whether to book a consultation. As a result, the agent saves admin time, while every judgement call stays with a person.

Accountants face many of the same questions, so our guide to AI agents for accountants is a useful comparison.

Confidentiality and security risks of AI agents for lawyers

Law firms hold exactly the kind of information attackers want. Therefore, AI agents for lawyers raise some specific risks:

• Data leakage: information typed into a free public chatbot falls under that service’s terms, which is why the Law Society says never to enter confidential data there.
• Prompt injection: an agent reviewing documents from the other side could meet hidden instructions in the text, as our prompt injection explainer describes.
• Over-broad access: an agent connected to a whole document system can see more than one matter needs.
• Invented authority: general tools can produce cases that look real but do not exist.

Our wider guide on whether AI agents are safe applies to firms as much as to consumers. Finally, be careful with vendor claims. Unless a product’s own documentation states that it meets a specific standard, do not assume it does, and even then treat it as the company’s statement.

What do AI agents for lawyers cost?

Public pricing for AI agents for lawyers is patchy as of October 2026. Thomson Reuters did not publish CoCounsel Legal pricing in its launch announcement. By contrast, Clio’s Grow AI uses a per-conversion model, which LawSites reported at $25 per converted lead, with three free conversions for existing Grow customers.

General assistants are usually cheaper to try, although business tiers with admin controls tend to cost more than personal plans. So check each vendor’s current pricing page before you commit.

Also budget for the hidden costs: staff time to review output, training, and the time a supervising lawyer spends checking work. The ABA’s fee guidance matters here too, because clients generally should not pay for a lawyer learning a tool.

Who are AI agents for lawyers for, and who should wait?

Firms with high volumes of similar documents, such as due diligence, employment claims or lease reviews, have the clearest use case for AI agents for lawyers. Similarly, firms that lose enquiries out of hours may benefit from an intake agent, as long as a lawyer reviews every lead.

However, it may be wiser to wait if your firm has no written AI policy or cannot get clear contract terms from a vendor. The same applies when you handle highly sensitive matters where any data exposure would be serious. Sole practitioners without time to review AI output should also go slowly, because the review is where the professional duty sits.

AI agents for lawyers: FAQs

Sources

• American Bar Association, “ABA issues first ethics guidance on a lawyer’s use of AI tools” (July 29, 2024)
• The Law Society, “Generative AI: a guide for solicitors” (updated September 2025)
• BIICL, “AI Fabricated Citations and the Legal Profession: Lessons from the High Court” (September 16, 2025)
• Thomson Reuters, “Thomson Reuters Launches Next Generation of CoCounsel Legal” (August 20, 2026)
• LawSites, “Clio Adds Agentic AI Capabilities to Clio Work, Also Launches Vincent Mobile App” (April 2, 2026)
• LawSites, “Clio Grows Clio Grow with Launch of Grow AI, Providing 24/7 Intake Agents” (August 20, 2026)
• Microsoft, “Introducing the new Copilot with Home, Code and Autopilot” (September 25, 2026)