Field note / Statistics
Legal document automation statistics (2026)
About 3 in 10 US private-practice attorneys said their offices used AI tools in 2024, about 1 in 5 legal professionals said their firm used generative AI firm-wide, and 1 in 5 measured its return. Preparing, filing or transmitting documents is the single largest activity behind legal malpractice claims, at over 35% of claims reported for 2020 to 2023. The figures below come from bar, academic, publisher and vendor sources, each linked and dated.
How many law firms use AI and document automation?
Adoption depends on what is counted. Individual use of any AI tool is high (79% in Clio's 2025 survey), while 30% of attorneys said their office used AI tools (ABA, 2024) and 21% of legal professionals said their firm used generative AI firm-wide (AffiniPay, 2024); both rise with firm size. The most recent public figure we found for document assembly software specifically is 36% of lawyers in 2020.
| Figure | What it measures | Source and year |
|---|---|---|
| 30% | US private-practice attorneys whose offices used AI-based tools (11% in 2023) | ABA Legal Technology Survey Report, 2024 data, via LawNext |
| 46% / 30% / 18% | AI tool use at firms of 100 or more attorneys / 10 to 49 attorneys / solos | ABA Legal Technology Survey Report, 2024 data, via LawNext |
| 36% | Lawyers who personally used document assembly software | ABA Legal Technology Survey Report, 2020 data, via MyCase |
| 21% firm-wide; 31% personal; 39% at firms of 51 or more lawyers | Generative AI use reported by legal professionals | AffiniPay Legal Industry Report 2025, 2024 data, 2,800 or more respondents, via the Federal Bar Association blog |
| 28% | Law firms actively using generative AI (26% across all legal respondents; 14% in 2024) | Thomson Reuters Institute, early 2025, 1,700 or more respondents (US, UK, Canada) |
| 79% | Legal professionals using AI in any capacity (same as 2024; 19% in 2023) | Clio Legal Trends Report 2025 and 2024, via 2Civility (2025) and 2Civility (2024) |
| 40% | Legal professionals using legal-specific AI tools (58% in 2024) | Clio Legal Trends Report 2025, via 2Civility |
| 68% | Law firm legal professionals using generative AI at least weekly (76% in corporate legal), 10 countries | Wolters Kluwer Future Ready Lawyer 2024, 712 respondents, via LawNext |
| 56% / 14% / 12% | US attorneys who had used generative AI / used it for work / whose firm had bought a tool, fall 2023 | Bloomberg Law survey, 450 or more attorneys, 2023 |
| 41% (UK) | UK and Ireland law firms using generative AI, September 2024 (11% in July 2023) | LexisNexis UK and Ireland survey, 800 or more respondents, via Local Government Lawyer |
How much lawyer time goes to drafting?
Public, independent measurement of drafting time is thin. The most cited figure, 40% to 60% of a lawyer's time, comes from Thomson Reuters' own research with no published sample. Clio's time-entry analysis gives a different lens: the share of billed work that could in principle be automated.
| Figure | What it measures | Source and year |
|---|---|---|
| 40% to 60% | Share of lawyers' time spent drafting documents | Thomson Reuters blog, June 2026; underlying research unnamed, no sample or data year stated |
| Often over 15 minutes | Time to find a starting point for a draft | Thomson Reuters blog, June 2026; underlying research unnamed, no sample or data year stated |
| 74% | Share of a law firm's hourly billable tasks that could be automated with AI (Clio reports 81% for administrative assistants, 69% for paralegals and 57% for lawyers); a modelled estimate from time entries | Clio Legal Trends Report 2024, via 2Civility |
How often do document errors occur, and what do they cost?
No public survey measures the error rate of law firm templates directly. The closest evidence is malpractice claims data, where preparing, filing or transmitting documents is the single largest activity behind claims, and contracting studies that put the cost of poor contract process at almost 9% of revenue. The hub field note shows how to test a document package against the cases that must stop it before anyone relies on the output.
| Figure | What it measures | Source and year |
|---|---|---|
| Over 35% | Share of legal malpractice claims arising from preparing, filing or transmitting documents (deeds, leases, contracts, wills, trusts) | ABA Profile of Legal Malpractice Claims 2020-2023, via ALPS Insurance |
| 51% | Share of alleged errors that were substantive errors (for example failure to know or apply the law) | ABA Profile of Legal Malpractice Claims 2020-2023, via ALPS Insurance |
| 13.58% | Share of all reported claims from estate, trust and probate practice | ABA Profile of Legal Malpractice Claims 2020-2023, via ALPS Insurance; the Swiss Re recap gives about 14% |
| Almost 9% | Average share of annual revenue lost to poor contract management (over 15% in complex industries) | World Commerce & Contracting, 2025 report, press release via CFO.com |
| 90% | Business users who find contracts difficult or impossible to understand | World Commerce & Contracting, 2025 report, press release via CFO.com |
| 69% | Organizations that do not require staff to use a template or pre-approved model when drafting a contract (31% follow a playbook; 71% of contracts are not monitored for deviations from standard terms) | EY Law and Harvard Law School Center on the Legal Profession, May 2021, 1,000 law department and contracting leaders |
| Over 52% (Australia) | Legal practitioners who admit skipping proofreading and checking because of time pressure | LexisNexis Australia, Drafting Research, June 2014, 650 or more respondents |
How reliable is generative AI output in legal work?
Benchmarks and court records both show a non-trivial error rate. Stanford's 2024 benchmark found leading legal research AI tools gave incorrect answers in 17% to 34% of queries, and a running database of court decisions dealing with hallucinated content passed 2,200 entries in October 2026, about 40% of them involving lawyers.
| Figure | What it measures | Source and year |
|---|---|---|
| More than 17% | Queries answered incorrectly by Lexis+ AI and Ask Practical Law AI (200 or more queries) | Stanford HAI and RegLab, May 2024; paper on arXiv |
| More than 34% | Queries answered incorrectly by Westlaw AI-Assisted Research | Stanford HAI and RegLab, May 2024 |
| 2,207 | Court decisions worldwide addressing AI-hallucinated content, as of 23 October 2026 (877 involving lawyers; 1,504 in the US; 234 in Canada; 76 in the UK) | Charlotin AI Hallucination Cases database, live |
| About 712 (estimated); about 90% in 2025 | Decisions in the same database at year-end 2025, and the share written in 2025 | Bloomberg Law, December 29, 2025 |
| 75% | US attorneys naming accuracy as their top AI concern (58% in 2023) | ABA Legal Technology Survey Report, 2024 data, via LawNext |
| 76% (UK) | UK legal professionals concerned about inaccurate or fabricated information | LexisNexis UK and Ireland survey, 2024, via Local Government Lawyer |
What do firms report after automating?
Self-reported savings cluster at 1 to 5 hours per week per AI user. Randomized trials with law students show speed gains of 12% to 32% with GPT-4 (2023) and productivity gains of 34% to 140% with a legal research tool and a reasoning model (2025). Only 20% of legal professionals measure return on investment, so most reported results are impressions rather than audited numbers.
| Figure | What it measures | Source and year |
|---|---|---|
| 65% / 12% / 7% | AI users saving 1 to 5 / 6 to 10 / 11 or more hours per week | AffiniPay Legal Industry Report 2025, 2024 data, via the Federal Bar Association blog |
| 61% / 21% | AI users reporting “somewhat” / significantly increased efficiency (2% reported a decrease) | AffiniPay Legal Industry Report 2025, 2024 data, via the Federal Bar Association blog |
| 54% | US attorneys naming “saving time/increasing efficiency” as the primary advantage of AI (44% in 2023) | ABA Legal Technology Survey Report, 2024 data, via LawNext |
| Nearly 3x | Likelihood that firms with wide AI adoption report revenue growth, compared with firms that have not adopted AI | Clio Legal Trends Report 2025, press release via CNW |
| 2x | Likelihood that growing firms use automation, compared with stable firms (nearly 3x compared with shrinking firms) | Clio Legal Trends Report 2025, press release via CNW |
| 5 hours per week; $19,000 per year | Predicted AI time savings per professional over the next year, and the stated annual value (all professions, 2,275 respondents) | Thomson Reuters Future of Professionals Report 2025, fieldwork February to March 2025 |
| 22%; 53% | Organizations with a visible, defined AI strategy; professionals saying their organization already sees a return on investment | Thomson Reuters Future of Professionals Report 2025 |
| 20% | Legal professionals measuring the return on investment of generative AI | Thomson Reuters Institute, early 2025 |
| 12% to 32% faster | Speed gain for law students using GPT-4 on legal tasks, randomized trial, 60 participants | Choi, Monahan and Schwarcz, 2023, via Legal Futures |
| 38% to 115%; 34% to 140% | Productivity gains with a retrieval-based legal AI tool and with a reasoning model, randomized controlled trial, 5 of 6 tasks | Schwarcz, Manning, Barry, Cleveland, Prescott and Rich, 2025, PDF hosted by UC Berkeley Law |
Want to know what one of your own workflows would save before building anything? Discuss one workflow in a short fit conversation; no documents needed.
How are law firms using generative AI?
Document work dominates. Among legal professionals who use generative AI, about three quarters use it for document review and summarization, and half or more for drafting briefs, memos and contracts. Policy lags use: roughly half of firms have no AI policy.
| Figure | What it measures | Source and year |
|---|---|---|
| 74% / 73% / 72% | Generative AI users applying it to document review / legal research / document summarization | Thomson Reuters Institute, early 2025 |
| 59% / 51% / 50% | Generative AI users applying it to brief or memo drafting / contract drafting / correspondence drafting | Thomson Reuters Institute, early 2025 |
| 72%; over 40% | Generative AI users who use it at least weekly; daily or more | Thomson Reuters Institute, early 2025 |
| 54% / 47% / 46% | AI users drafting correspondence / brainstorming / doing general research (45% use AI daily, 40% weekly) | AffiniPay Legal Industry Report 2025, 2024 data, via the Federal Bar Association blog |
| 65% / 56% | Lawyers naming legal research / drafting documents as the tasks where generative AI could help (3,752 lawyers, US, UK, Canada, France) | LexisNexis International Legal Generative AI Survey, March to July 2023 |
| 53%; 52% | Legal professionals whose firm has no AI policy or who are unaware of one (Clio); organizations with no generative AI usage policies (Thomson Reuters) | Clio Legal Trends Report 2025, via 2Civility; Thomson Reuters Institute, 2025 |
| 35%; 78% | Legal professionals who rarely or never disclose AI use to clients; clients who want disclosure | Clio Legal Trends Report 2025, via 2Civility |
How to cite this page
Stratum Relay, “Legal document automation statistics (2026),” www.stratumrelay.com/field-notes/legal-document-automation-statistics, reviewed . Each figure above should be cited to its original source, which is linked in the table; this page is a curated index, not the origin of the data.
Method and limits
- Source order of preference: bar associations and insurers' claims data, academic studies, publisher research institutes (Thomson Reuters, Bloomberg Law, LexisNexis, Wolters Kluwer), then vendor surveys that state their sample size and date. Every figure was read on the linked page on 22 or 23 October 2026; nothing is quoted from memory.
- “Via” rows: where a publisher's own page could not be retrieved (the ABA TechReport, Clio's report site, Wolters Kluwer's site), the table links to the summary page that was read and names the original report.
- Vendor surveys (Clio, AffiniPay, Thomson Reuters, LexisNexis, Wolters Kluwer) sample their own customers or opt-in panels and have an interest in reporting high adoption and large savings. Their figures are reported as stated and should not be combined with each other or with the ABA figures; each survey defines “use of AI” differently.
- Years: the year shown is the year of the data where the page states it; otherwise it is the publication date. Several figures are from 2020 to 2023 and are kept only because no newer public figure was found. The 40% to 60% drafting-time figure has no sample, data year or report name on its page and is included because it is the figure most often repeated; treat it as weak.
- Estimates versus measurements: “could be automated” (Clio) is a modelled estimate, not a measurement. Hours-saved figures are self-reported predictions or impressions, except the two randomized trials, which used law students rather than practising lawyers.
- Claims data covers claims reported to participating insurers, not all errors, and does not isolate template or precedent errors.
- This page is general information about the legal technology market. It is not legal advice and does not describe any client's matter.
Please do not send confidential documents or client records. A conversation can start with the document package, who owns it and the tools you already use.
Sources
All pages accessed 22 or 23 October 2026.
- LawNext, ABA tech survey finds growing adoption of AI in legal practice (March 2025), reporting the ABA 2024 Legal Technology Survey Report.
- MyCase, ABA survey results: lawyers and cloud computing in 2020, reporting the ABA 2020 Legal Technology Survey Report.
- 2Civility (Illinois Supreme Court Commission on Professionalism), 2025 Clio Legal Trends Report (October 2025).
- 2Civility, Clio 2024 Legal Trends Report (October 2024).
- Clio press release via CNW, Clio's 2025 Legal Trends Report (October 2025).
- Thomson Reuters Institute, Generative AI in Professional Services: executive summary for legal professionals (April 2025).
- Thomson Reuters press release, Future of Professionals Report 2025 (June 2025).
- Thomson Reuters, Legal drafting: challenges, risks and opportunities (June 2026).
- Federal Bar Association blog (sponsored post by LawPay), The Legal Industry Report 2025 (April 2025), presenting the AffiniPay report.
- LawNext, Future Ready Lawyer survey from Wolters Kluwer (October 2024).
- Local Government Lawyer, Most lawyers now using generative AI or plan to, survey says (September 2024), reporting a LexisNexis UK and Ireland survey.
- Bloomberg Law, AI fever in legal industry cools off, survey suggests (November 2023).
- ALPS Insurance, How to manage the risks associated with substantive malpractice missteps (2025), summarizing the ABA Profile of Legal Malpractice Claims 2020-2023.
- Swiss Re Corporate Solutions, ABA Profile of Legal Malpractice Claims (November 2024).
- World Commerce & Contracting press release via CFO.com, The multi-million dollar leak in your contracts (August 2025).
- EY Law and Harvard Law School Center on the Legal Profession, How does contracting complexity hide clear profitability? (May 2021).
- LexisNexis Australia, Drafting research (undated; survey June 2014).
- Stanford HAI and RegLab, AI on trial: legal models hallucinate in 1 out of 6 (or more) benchmarking queries (May 2024); paper: arXiv 2405.20362.
- Damien Charlotin, AI Hallucination Cases database (live; read 19 October 2026 update).
- Bloomberg Law, AI-faked cases become core issue irritating overworked judges (December 2025).
- Legal Futures, AI helps lawyers complete work quicker, study finds (November 2023), reporting Choi, Monahan and Schwarcz.
- Schwarcz, Manning, Barry, Cleveland, Prescott and Rich, AI-Powered Lawyering (2025, PDF hosted by UC Berkeley Law).
- LexisNexis, International Legal Generative AI Survey (August 2023).
For practice context, see how Stratum Relay works. This organizational attribution is not a statement that a lawyer, client or named individual has reviewed or endorsed the page.