Where AI helps
Your product is judgement, put into documents. AI is good at producing documents. It is unreliable at judgement and at facts. So the gain depends on the task, and on who checks.
| Work | What has been shown | Where it comes from |
|---|---|---|
| Drafting and writing | Professionals finished writing tasks 40% faster, with better quality. | Experiment with 453 professionals |
| Analysis and client deliverables | Consultants did more and better work on 18 tasks. On one task that looked similar, they were more often wrong with AI than without. | Experiment with 758 consultants |
| Research and citations | A general chatbot invented six court decisions. A consulting report was reissued to fix wrong citations. A second Deloitte report, in Canada, cited research papers that do not exist. Such cases are not rare: one researcher’s database lists 2,199 court decisions involving AI-invented material, 871 of them involving lawyers, as of 16 October 2026. | Law firm case, consulting case, HR Reporter on Deloitte Canada, the AI Hallucination Cases database |
| An AI-run law firm | England’s legal regulator authorised Garfield.Law in May 2025. The firm uses AI to recover unpaid debts for small firms through the small-claims court. Its conditions: the AI may not propose case law, it acts only when the client approves each step, and named solicitors stay accountable. | Garfield.Law case, from the regulator’s decision |
| Routine processing | At one accounting firm, years of automated processing wore down staff skills. That was older software, not today’s AI. | Case study of an accounting firm |
Use is wide. Gains are uneven. In a BCG survey of US professional-services workers, 95% used generative AI at least once a month. Yet 53% said it saved them less than a quarter of their time. Among US lawyers, 30% reported using AI tools in the American Bar Association’s 2024 survey, up from 11% a year earlier. In a 2026 survey by Clio, which sells legal software, 71% of solo practitioners and 75% of small law firms used AI for legal work. Fewer than a third had grown revenue with it, and 86% of solo practitioners and 78% of small firms had made no pricing change because of AI.
Smaller firms start further back. A study of several audit firms found that smaller firms were less ready for AI than large ones.
Risks and rules
- The firm answers for the output. In the law firm case, a US federal court said lawyers keep “a gatekeeping role” over the accuracy of what they file, whatever tool they use. The firm was fined.
- Client confidentiality. In the United States, the American Bar Association issued guidance for lawyers in July 2024. It says a lawyer using generative AI must keep client information confidential unless the client gives informed consent. Informed consent means the client agrees after the risks are explained. The same guidance covers competence, telling clients how the work is done, and fees.
- Fees. The same guidance says a lawyer may charge for time spent giving a tool its input and checking the draft. In most cases the lawyer may not charge the client for learning to use the tool. If you bill by the hour, an hour the tool saves is an hour you do not bill, unless you change how you price. See measure time saved against a baseline.
- Leaders answer for the checks. In June 2025 the High Court of England and Wales said that people who lead legal practices, “such as heads of chambers and managing partners”, must take “practical and effective measures” so that everyone understands the duty to check AI output. Publishing guidance is not enough by itself. See the Primus Solicitors case, from the same judgment.
- Staff must understand the tools. In the EU, the AI Act requires organizations that use AI to “take measures to support the development of AI literacy of their staff”. AI literacy means knowing enough about AI to use it sensibly. The duty has applied since 2 February 2025. It does not require a firm to guarantee any set level of knowledge. See test AI knowledge instead of asking staff to rate their own.
- Junior staff learn on the tasks AI takes. First drafts, research and checking are how juniors build judgement. Whether AI is already cutting junior hiring is disputed. In US résumé data, firms that advertised jobs for people to build generative AI into their work hired fewer juniors afterwards, and 28% of those firms were in professional services (study). Other US data find no clear disruption across the labour market. See keep the entry-level tasks that newcomers learn on.
Practices for this industry
All 30 practices in the library apply in this industry. None of them is specific to one sector. For the ones below, a study, a company case or a worked example from this sector shows why they matter here.
- Keep a person checking anything that goes to a customer or carries legal weight Five documented cases in law and consulting involve made-up or wrong citations that nobody looked up before they went out.
- Train staff on the tools you actually chose A lawyer who had never used ChatGPT took it for a search engine and filed six court decisions ChatGPT had made up.
- Budget for the work AI creates Checking sources is the new work. Two documented failures, a court filing and a consulting report, involved citations nobody had checked.
- Know the rules that already apply to your AI use Courts in the United States and England have applied lawyers’ existing duties to AI-assisted work and fined or referred those who filed unchecked output.
- Write a one-page AI-use policy staff will actually read One law firm’s policy required written approval to use ChatGPT for research. A partner ignored it and filed five made-up citations.
- For each kind of task, test whether a person with AI beats either one alone Tested on 758 management consultants. AI helped on most tasks and hurt on one that looked similar.
- Test AI knowledge instead of asking staff to rate their own A lawyer who took ChatGPT for a search engine filed six court decisions it had made up.
- Claim only what your AI has been tested to do A service sold as a “robot lawyer” had never been tested against a lawyer. Two investment advisers were fined for AI they did not have.
- Put your own records in order before you ask AI to predict from them A two-person advisory firm names one file per client, in place of scattered files, as a choice its AI project depended on.
- When AI output goes wrong, correct it fast and in the open Courts in three cases said in terms that a prompt, honest admission reduced the penalty, or that defending the error made it worse.
- Set the limits before the tool acts A two-person advisory firm lets AI send reminders unaided. The adviser approves important letters.
- Keep the entry-level tasks that newcomers learn on At one accounting firm, years of relying on automated software wore down staff skills, and nobody noticed.
Pick your company size
Size changes what is practical more than sector does. Each link opens the library filtered to that size.
Company cases
Each case is a named company with numbers and a public source. All cases.
- Ashurst: first drafts 45–80% faster in timed trials, with lawyers still ahead on accuracy The law firm Ashurst tested generative AI with 411 of its people for five months and published the results. First drafts came 45% to 80% faster in timed tests. In a blind comparison, lawyers’ own drafts were more accurate on average, and one AI draft cited a wrong verdict.
- Butler Snow: a firm with an AI policy, and a partner who did not follow it The firm’s policy required written approval before using ChatGPT for legal research. A partner used it without approval and filed five citations it had made up. The court removed three lawyers from the case. The firm paid for a check of 2,400 citations.
- Deloitte Australia: a government report reissued to fix wrong citations A Deloitte report for an Australian government department had to be reissued to correct citations and a summary of a court case. The corrected version disclosed the use of a generative AI tool. The report does not say AI caused the errors.
- DoNotPay: a “robot lawyer” that was never tested against a lawyer DoNotPay told customers its AI could do the work of a lawyer. The US consumer regulator said the company had never tested whether that was true and employed no lawyers to check. It paid $193,000 and had to tell past subscribers about the limits.
- Garfield.Law: a regulator authorised an AI-run law firm for small debt claims, with limits The regulator of solicitors in England and Wales authorised its first law firm that delivers legal services through AI. It helps businesses recover unpaid debts through the small-claims court. The regulator set limits: the AI may not propose case law, takes no step the client has not approved, and named solicitors stay accountable.
- Gilberts: a smaller audit firm started with the AI in its existing tools and two firm rules A smaller UK audit and accountancy firm began with the AI features in tools it already had, on low-risk internal tasks. A partner reports that summarising a long document now takes 10 to 15 minutes where it took 45. The firm bars client data from public AI tools and treats every AI output as a first draft.
- Levidow, Levidow & Oberman: fined after filing court papers with cases ChatGPT made up A lawyer used ChatGPT for legal research and filed six court decisions that did not exist. The court fined the firm and two of its lawyers. The judge said using AI was not the problem. Not checking was.
- Morgan & Morgan: eight made-up cases from the firm’s own AI tool, signed by three lawyers A lawyer asked his firm’s own AI tool to add case law to a court filing. Eight of the nine cases it cited did not exist. Two other lawyers signed without reading. The firm had trained staff, and it then added a warning inside the tool.
- Primus Solicitors: 18 court citations that did not exist, taken from a client’s AI research A solicitor at a small law firm filed a statement with 45 case citations. Eighteen of the cases did not exist. His client had produced the research with public AI tools, and nobody at the firm checked it. The court referred him to his regulator.
- SAM: a two-person advisory firm cut the paperwork on each client file from six hours to two and a half A two-person firm that advises clients on retirement turned each client file into a fixed sequence of steps and gave the routine ones to AI. Administrative time per file fell from about six hours to two and a half. Set-up cost about €3,000 and 60 hours, and the adviser still approves important letters.
Studies from this sector
These studies were run on work that is typical of this sector.
- The jagged frontier: AI helped consultants on some tasks and hurt on another 758 consultants at Boston Consulting Group. With AI they did more, faster and better on 18 tasks. On one task beyond the AI’s ability they were 19% less likely to be right.
- ChatGPT and professional writing tasks: an experiment 453 professionals on writing tasks typical of their jobs. With ChatGPT they finished 40% faster and quality rose by 18%.
- How automation eroded skills at an accounting firm One accounting firm. Years of relying on automated software wore down staff skills, and nobody noticed until it mattered.
Sources
- BCG (2025). Gen AI in professional services. Survey of US professional-services workers, run in late 2024. bcg.com
- LawNext (March 2025), reporting the American Bar Association’s 2024 technology survey. lawnext.com
- Yang, Blount and Amrollahi (2024). Artificial intelligence adoption in a professional service industry: A multiple case study. Technological Forecasting and Social Change. doi.org
- American Bar Association (29 July 2024). ABA issues first ethics guidance on a lawyer’s use of AI tools. Announcement of Formal Opinion 512. americanbar.org
- European Union. AI Act, Regulation (EU) 2024/1689, Article 4 (AI literacy), as amended in 2026. Text on the European Commission’s AI Act Service Desk. ai-act-service-desk.ec.europa.eu
- Charlotin. AI Hallucination Cases database. Count as of 16 October 2026. damiencharlotin.com
- Clio (4 May 2026). Legal Trends for Solo and Small Law Firms 2026. Press release by a legal software vendor. clio.com
- Hosseini Maasoum and Lichtinger (May 2026). Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data. SSRN 5425555. Working paper. raw.githack.com
- Solicitors Regulation Authority (May 2025). Authorisation of Garfield.Law. sra.org.uk
- R (Ayinde) v London Borough of Haringey and Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin). Judgment of 6 June 2025. High Court of England and Wales, Divisional Court. judiciary.uk