Legal AI: what is getting real adoption now?
In our Legal Tech market deck, you will find everything you need to understand the market
SUMMARY
Legal AI is getting real adoption now in research, document analysis, contract work, drafting and due diligence, with supervised multi-step agents beginning to join that group.
The clearest sign that legal AI has moved beyond pilots is repeated use. Organizational adoption is rising, weekly usage is common, and large firms are putting legal AI into production workflows rather than limiting access to innovation teams.
Adoption is clustering around work that lawyers can verify. Research answers can be checked against authorities, document summaries against the underlying file, and generated clauses against precedent, which makes these tasks much easier to trust on live matters.
Legal research appears to have the broadest reach, but contract review may be producing the clearest business case. Contract teams can measure review time, turnaround speed, stakeholder response and outside-counsel costs much more easily than they can measure the value of a better research session.
Drafting adoption is real, although the dominant workflow is first-draft compression rather than autonomous authorship. Lawyers increasingly use AI to get from a blank page to something worth editing, especially when the system can draw on firm precedents and matter knowledge.
The biggest product shift is happening behind the chatbot. Legal AI becomes much more useful once it sits directly inside document repositories, Word, research databases, eDiscovery systems, contract platforms and firm knowledge rather than forcing lawyers to move information manually between tools.
General-purpose models have won enormous reach, while specialist legal platforms are gaining ground where privileged files, authoritative law, permissions and firm-specific knowledge matter. The boundary between those two categories is already getting blurrier as the large model providers connect directly to legal systems.
Small firms have largely caught up in basic AI usage. Big Law still has an advantage in integration because it can spend more on security, knowledge systems, training and practice-specific workflows, so the difference is increasingly about depth rather than access.
Legal agents are useful now, but autonomy remains much weaker than the marketing language suggests. Multi-step research, diligence and analysis workflows can already remove a lot of manual prompting, yet current benchmark results still make full matter-level delegation hard to justify.
The next adoption fight may be commercial rather than technical. Once clients know that research, drafting and review can be completed much faster, hourly billing becomes harder to defend and firms have to decide who captures the productivity gain.
The products gaining the strongest foothold are therefore the ones that make lawyers much faster without pushing the lawyer too far away from the evidence. Legal AI has already taken meaningful pieces of the workflow; strategy, final judgment and responsibility for an entire matter remain much less automated.
Has legal AI actually moved beyond pilots?
As of now, legal AI has clearly moved beyond pilots for research, document review, drafting and contract work.
The best evidence is repeated use on real work rather than the number of firms announcing AI partnerships. Thomson Reuters’ 2026 AI in Professional Services survey found that 41% of law firms were already using generative AI at an organizational level, up from 28% a year earlier. Corporate legal departments moved from 23% to 47%. Among professionals already using generative AI, 82% said they used it at least weekly.
LexisNexis’ latest 2026 survey points even further in the same direction. Of 543 legal professionals surveyed, 94% said they now use AI for legal work and 74% use it at least once a week. Different surveys define “use” differently, so 94% should not be treated as the universal adoption rate for every lawyer. The useful part is the frequency: a large share of lawyers have moved well beyond occasionally testing a chatbot.
Large-firm production data makes the picture harder to dismiss. The 2026 SKILLS Legal AI survey asked AI and innovation leaders at 130 of the world’s largest law firms which tools were actually live. More than 40 firms reported production deployments in categories such as legal drafting, contract review, due diligence and eDiscovery.
We also have evidence that some deployments survive the novelty phase. Harvey currently reports a 92% monthly adoption rate across surveyed customers. During Maddocks’ pilot across 13 practice teams, 70% of participating lawyers ran queries daily and 88% came back week after week before the firm decided on an enterprise-wide rollout.
For the rest of this article, we will call legal AI genuinely adopted when lawyers keep using it on live work after the security review, training and initial excitement are over. By that standard, several legal AI workflows are already real.
If you want more recent data on this point, please see our latest Legal Tech market report.
What are lawyers actually using legal AI for right now?
Legal AI adoption today is heavily concentrated in research, reading documents and producing first drafts.
Thomson Reuters asked legal professionals already using generative AI what they were doing with it. Legal research led at 80%. Document review reached 74%, document summarization 73%, brief or memo drafting 59%, correspondence drafting 55% and contract drafting 49%.
The order tells us quite a lot. Lawyers first gave AI work that is frequent, time-consuming and easy to inspect afterward. A research answer can be checked against the cited cases. A summary can be checked against the document. A generated clause can be compared with the firm’s preferred language.
The further AI moves toward making the final legal call, the thinner the adoption evidence becomes. Current models are getting plenty of substantive work, but lawyers are still keeping the final decision.
| Legal AI task | Share of current GenAI users in Thomson Reuters’ 2026 survey | Where adoption stands |
|---|---|---|
| Legal research | 80% | Already mainstream among AI users |
| Document review | 74% | Already mainstream among AI users |
| Document summarization | 73% | Already mainstream among AI users |
| Brief or memo drafting | 59% | Common with lawyer review |
| Correspondence drafting | 55% | Common |
| Contract drafting | 49% | Established, especially in transactional work |
This market map, featured in our Legal Tech market deck, highlights top companies and startups in the Legal Tech market
Is legal research still the biggest legal AI use case?
Legal research is currently the broadest legal AI use case we can identify across independent surveys.
LexisNexis found research among the leading legal AI workflows in its surveys, while the State Bar of Texas reached a similar result from a very different sample. Its 2026 survey found that attorney AI use had risen from 30% in 2024 to 62%, and legal research was the most common use case at 53%.
Research works unusually well as an entry point because lawyers do it constantly and can verify the answer. The important change lately is that lawyers increasingly care about what sits behind the answer. In LexisNexis’ newest survey, 81% said they felt more comfortable using AI when it was grounded in trusted legal sources, up from 70% a year earlier.
That helps explain why Westlaw, Lexis and newer platforms can keep growing even when lawyers already have access to ChatGPT or Claude. Asking a general model to suggest where to look is useful. Asking a system to search authoritative law, show the sources and let the lawyer inspect them is much easier to use on client work.
Research probably wins on breadth rather than spectacular time savings. A diligence review involving 600 agreements can produce a more dramatic productivity jump, but research touches far more lawyers, practice areas and matters every day.
Are lawyers now using AI to review and summarize documents every day?
AI document review and summarization are already ordinary legal AI workflows in many firms.
The usage numbers sit just behind research. In Thomson Reuters’ survey, roughly three-quarters of current legal GenAI users were reviewing or summarizing documents with AI. Other legal-industry surveys place both tasks among the most common uses as well.
The appeal becomes obvious when the document set gets large. A lawyer gains little strategic advantage from manually reading the same boilerplate provision 200 times. AI can pull out change-of-control clauses, renewal dates, termination rights or unusual liability wording across the whole set and give the lawyer a smaller group of exceptions to inspect closely.
The technology is also moving deeper into the systems where those documents already live. Thomson Reuters’ latest CoCounsel Legal release connects legal analysis and drafting more closely to large document collections, including tabular analysis across as many as 10,000 documents. Its forthcoming Reveal integration is designed to let litigation teams move reviewed evidence into CoCounsel without repeatedly exporting and uploading files.
That direction is more useful than another chatbot feature. Document AI gets much better when the lawyer can ask questions against the actual matter file, see where the answer came from and immediately open the source.
As this chart shows, and as featured in our Legal Tech market deck, search interest in Legal Tech has been growing steadily
Are lawyers actually using AI to draft legal documents?
Lawyers are already using AI heavily for first drafts, rewrites and clause work, while final legal drafting still stays under human review.
More than half of legal GenAI users in Thomson Reuters’ survey used AI for briefs or memos, and almost half used it for contract drafting. The 2026 Legal Industry Report found a similar pattern for specialist legal AI, with document drafting used by 49% of respondents who used those tools.
The everyday workflow is usually less dramatic than “write this entire merger agreement.” Lawyers ask AI to produce a first pass of a memo, turn notes into a structured argument, rewrite a clause, compare language with precedent, draft a response or convert research into a usable document.
Products are increasingly being designed around that behavior. Spellbook works inside Microsoft Word rather than asking transactional lawyers to move their drafting into a separate chatbot. Harvey and Legora also push drafting directly into document and firm-knowledge workflows.
Freshfields has gone further by signing a multi-year agreement with Anthropic to deploy Claude globally and co-build legal workflows. Kirkland & Ellis is spending heavily on its own AI infrastructure and has built a fund-formation platform with Palantir that puts firm knowledge into workflows used by more than 1,000 lawyers in its Investment Funds Group.
The adoption story here is first-draft compression. Lawyers can get to something worth editing much faster, particularly when the AI can work from the firm’s own precedents.
Is contract review where legal AI is paying off fastest?
Contract review currently gives us some of the clearest evidence that legal AI is producing measurable business value.
LegalOn and In-House Connect surveyed 452 in-house legal professionals for their 2026 report. Fifty-two percent were already using or evaluating AI for contract review, and active usage had nearly quadrupled compared with 2024. Respondents said reviewing one contract took an average of 3.1 hours, which leaves a large and repeated chunk of work for AI to compress.
Ironclad’s separate 2026 survey of 822 legal professionals reached an even stronger conclusion about results. It ranked contract review as the most impactful AI use case. Among respondents reporting measurable outcomes, 52% said AI helped them respond faster to business stakeholders, 50% saw faster contract turnaround and 42% reduced outside-counsel spending.
Those figures come from vendors that sell contract technology, so the exact percentages deserve some caution. The broader pattern is harder to argue with because the same workflow keeps appearing across surveys and deployments.
Contract review also gives AI unusually clear instructions. A company can provide its standard positions on indemnities, liability caps, governing law, renewal terms and dozens of other clauses. The AI then compares incoming contracts with those positions, highlights deviations and suggests fallback language.
That makes contracting unusually executable. The system has a playbook, a document and a set of deviations to find, which helps explain why this part of legal AI has moved so quickly.
| Contract AI evidence | Result |
|---|---|
| In-house teams using or evaluating AI contract review | 52% |
| Average time previously spent reviewing one contract | 3.1 hours |
| Ironclad respondents seeing faster stakeholder response | 52% |
| Ironclad respondents seeing faster contract turnaround | 50% |
| Ironclad respondents reporting lower outside-counsel spend | 42% |
If you want more recent data on this point, please see our latest Legal Tech market report.
This chart, featured in our Legal Tech market deck, illustrates yearly venture capital funding for Legal Tech startups
Is legal AI actually being used for M&A due diligence?
AI is now being used in production for M&A and other due diligence, especially for the first review of large document sets.
The 2026 SKILLS survey found Harvey live for due diligence at 39 of the 130 large firms surveyed. Contract review and contract negotiation were each live at 41. These were production deployments rather than tools sitting on a trial list.
Due diligence is a natural fit because so much of the work starts with finding and classifying things. A deal team may need every agreement containing a change-of-control provision, unusual termination right, exclusivity clause or customer consent requirement. AI can search the full population, extract the relevant language and organize the findings before a lawyer decides what could actually affect the deal.
The scale changes the economics. When reviewing another hundred agreements becomes cheap, teams can inspect a larger share of the data rather than relying as heavily on sampling. They can also rerun questions when the scope of the deal changes.
AI is taking over a lot of the searching, extracting and first-pass comparison. The lawyer remains responsible for deciding whether an unusual clause is material, how it affects negotiations and what the client should do about it.
Are litigators using legal AI for more than research?
Litigators are now using legal AI for evidence review, chronologies, deposition preparation and drafting as well as research.
Some of this market predates generative AI. Litigation teams have used machine learning and technology-assisted review in eDiscovery for years. The newer step is putting conversational and generative systems on top of that evidence.
Instead of receiving only a relevance score, lawyers can ask questions across a case file, extract factual timelines, compare witness accounts, find inconsistencies and turn those findings into deposition questions or draft sections of a brief.
Current product development is following exactly that path. Harvey has been integrating more closely with litigation and docket systems, including Everlaw and PacerPro. Thomson Reuters is connecting CoCounsel Legal with Reveal so reviewed evidence can flow directly into research, analysis and drafting.
This is likely to be one of the more durable forms of litigation AI. A model answering from a controlled evidence set gives the lawyer something concrete to verify. Asking AI to understand the entire theory of a case, decide which facts really matter and choose the litigation strategy is a much bigger jump, and we have far less evidence of lawyers handing over that level of judgment.
This chart, featured in our Legal Tech market deck, looks at Clio’s strategy in Legal Tech
Are lawyers using ChatGPT or specialist legal AI?
Lawyers currently use both general AI and specialist legal AI, with general tools winning on reach and legal platforms winning more high-stakes workflows.
The State Bar of Texas gives us a simple illustration. Among attorneys who used AI, 63% used ChatGPT, making it the most common tool in the survey. Westlaw or CoCounsel was the leading specialist legal option at 30%.
The 2026 Legal Industry Report found the same split at a larger scale. Sixty-nine percent of more than 1,300 surveyed legal professionals personally used general-purpose AI for work. At the firm level, 46% had implemented general AI and 34% had implemented legal-specific AI.
The task mix changes when specialist tools are available. The same report found 58% using legal-specific AI for legal research, compared with 38% using general AI for research. Specialist legal AI was also widely used for drafting, summarization and other document-heavy work.
The newest LexisNexis data helps explain the gap. As discussed earlier, 81% of surveyed lawyers now say they are more comfortable when AI answers are grounded in trusted legal sources. Confidential client documents, citation checking, firm precedents and matter permissions create similar pressure.
ChatGPT and Claude remain very attractive for brainstorming, rewriting, correspondence and general questions. Once the work depends heavily on privileged files, authoritative case law or a firm’s own knowledge, specialist infrastructure becomes far more useful.
The boundary could become less visible soon. Google recently launched Gemini Enterprise for Legal in preview, while OpenAI has been making a much bigger push into legal software integrations. General AI companies increasingly want their models connected directly to the same databases and workflows that specialist vendors built their businesses around.
If you want more recent data on this point, please see our latest Legal Tech market report.
Have small law firms caught up with Big Law on AI?
Small law firms have caught up surprisingly fast in basic AI use, although Big Law is still further ahead in deep, firm-wide integration.
Clio’s 2026 research found that 71% of solo practitioners and 75% of small firms were using AI for legal work. Mid-sized firms were even higher at 86%. Those numbers are miles away from the picture of AI as something only giant firms can afford.
The remaining gap appears when we ask how deeply AI changes the business. Clio found that only around a third of solo and small firms had increased revenue after adopting AI, compared with 59% of enterprise firms. Many smaller firms are getting faster without redesigning pricing, intake, staffing or the amount of work they can handle.
Large firms can also build much heavier infrastructure around the models. CMS reached more than 95% active adoption among the first 3,000-plus lawyers given Harvey before expanding access across its 7,200-lawyer network. Baker McKenzie has rolled Legora out across its global network. BCLP made Legora available to 1,200 lawyers after testing competing platforms. Davis Wright Tremaine recently expanded firm-wide access to Harvey alongside Microsoft’s AI tools.
The gap today is less about whether a lawyer has tried AI. Big Law has more money and internal expertise to connect AI with document management, precedents, security controls, training and practice-specific workflows. That gives the technology more chances to become part of how the firm actually operates.
This chart, featured in our Legal Tech market deck, illustrates yearly funding for Legal Tech startups
Are in-house legal teams adopting AI faster than law firms?
In-house legal teams are currently moving at least as fast as law firms, and contract-heavy departments may already be ahead.
Thomson Reuters found 47% of corporate legal departments using generative AI at the organizational level versus 41% of law firms. As we saw earlier, the corporate figure had jumped from 23% the previous year.
Deloitte reached the same general conclusion from a different sample of senior legal leaders. Its 2026 research found 61% of legal departments had entered AI deployment phases, while 10% said AI was already embedded across daily workflows.
Another LegalOn study shows why those headline adoption numbers need context. Its 2026 In-House Legal Pulse found 92% of teams using AI somewhere, but only 11% had reached what it classified as connected or full-workflow maturity. Legal departments are clearly using the technology much faster than they are rebuilding entire processes around it.
The pressure is easy to see. In-house lawyers repeatedly handle NDAs, vendor contracts, procurement requests, employment questions, compliance work and internal legal requests. If a contract can be reviewed faster, the effect is visible to sales or procurement immediately. Saving outside-counsel costs makes the ROI even easier to explain.
Corporate legal departments therefore have a very practical reason to push AI beyond experimentation: the rest of the company is waiting for their answers.
Is Harvey already winning legal AI?
Harvey is currently the strongest pure-play legal AI platform by visible scale, but the market is still far too competitive to call it won.
Harvey says more than 200,000 lawyers across 2,400 organizations in 70 countries use its platform. Its systems now process more than 850,000 queries a day and over 50 million files a day. Those are vendor-reported figures, but the volume is large enough to show that legal AI has become serious enterprise software rather than a collection of small trials.
Legora has scaled quickly as well. Its current newsroom says the platform is used by more than 100,000 legal professionals at more than 1,500 law firms and in-house teams across over 50 markets. Earlier this year, Legora said it had crossed $100 million in annual recurring revenue less than 18 months after its general launch.
The incumbents remain enormous. CoCounsel has been chosen by one million professionals across 107 countries and territories, although that figure includes tax, audit, accounting, compliance and other professions alongside legal. Thomson Reuters says CoCounsel reaches 90 of the Am Law 100.
Then Google entered the market. Gemini Enterprise for Legal recently launched in preview with Cleary Gottlieb, Freshfields, Weil and Williams & Connolly among the launch customers. Its architecture looks strikingly similar to where the whole legal AI market is heading: specialist legal skills, agents, firm data, legal databases and connectors into document, contract and eDiscovery systems.
Some giant firms are building their own layer as well. Kirkland & Ellis has set aside $500 million to develop proprietary AI technology while continuing to pay for third-party products. Its chairman has explicitly framed the goal around making the firm’s collective knowledge available throughout the organization.
As we saw above, the SKILLS survey also found several tools living side by side in the same use cases. Kira remained widely deployed for contract review while Westlaw products remained heavily used for research even as Harvey expanded.
The market increasingly looks like a fight over who becomes the main interface through which lawyers reach their research databases, documents, precedents, workflows and models. Harvey has an impressive lead among the AI-native vendors, but Westlaw, Lexis, Legora, Google, OpenAI and firms building their own systems all have credible ways into that layer.
| Legal AI platform | Recent scale evidence | What the number actually tells us |
|---|---|---|
| Harvey | 200,000+ lawyers; 2,400 organizations | Largest visible pure-play deployment footprint |
| Legora | 100,000+ legal professionals; 1,500+ organizations | Fast-growing direct challenger |
| CoCounsel | 1M professionals; 107 countries and territories | Huge professional installed base, broader than legal alone |
| Gemini Enterprise for Legal | Preview with major global firms | Too early to call adoption, but a serious new entrant |
| Kirkland proprietary AI | $500M set aside for development | Large firms may build their own intelligence layer too |
If you want more recent data on this point, please see our latest Legal Tech market report.
This chart, featured in our Legal Tech market deck, compares the main business model options for Legal Tech SaaS platforms
Can legal AI agents actually run a legal matter yet?
Legal AI agents can now handle useful multi-step jobs, but they still cannot reliably run a complex legal matter from beginning to end.
This is where the gap between product marketing and demonstrated capability is widest. LegalOn found 80% of surveyed in-house teams exploring or evaluating AI agents, while respondents strongly preferred supervised automation. Firms clearly want agents, but wanting them and trusting them with an entire matter are very different stages of adoption.
Harvey’s Legal Agent Benchmark gives us a useful reality check. The benchmark was built around long legal assignments with matter files, instructions and a finished work product. In its initial 2026 results, frontier models completed fewer than 10% of tasks end-to-end under the benchmark’s strict all-pass scoring rule.
That scoring standard is deliberately hard: one missed requirement can fail the whole task. Even with that caveat, less than 10% is nowhere near the reliability needed to hand an agent a complicated matter and stop checking what it does.
The products are still becoming more agentic. Legora says customers increasingly build multi-step workflows. Harvey has more than 25,000 custom Workflow Agents across customer organizations. The latest CoCounsel Legal is designed to turn a plain-language request into a longer sequence of research, analysis and drafting. Google is explicitly pitching specialized legal agents.
Those tools can already save lawyers from manually moving through five or ten individual prompts. A diligence agent might review documents, classify provisions and produce a structured report. A research workflow might gather authority, compare it and create a draft memo.
Running the matter requires much more. The system would need to recognize what it has misunderstood, notice missing information nobody told it to look for, decide which legal and commercial risks deserve attention, change strategy when the facts move and know when to ask the client a question. Current adoption does not show lawyers routinely trusting AI with that job.
For now, legal agents are becoming good workflow executors under supervision. Full autonomous lawyering is still ahead of the evidence.
Is legal AI already forcing law firms to change prices?
AI is now putting real pressure on law firm pricing, although billing models are changing much more slowly than the technology.
Thomson Reuters found 71% of in-house legal professionals expecting outside firms to change their commercial models as AI usage increases. Only 28% of law firms said they had actually changed pricing in response.
The pressure has become much more concrete lately. Morgan Stanley, Citigroup and Goldman Sachs have been among the large financial clients pushing outside law firms to show how AI efficiencies affect what clients pay. Citi and Morgan Stanley have been moving more work toward competitive bidding and alternative fee arrangements, while Goldman has asked firms for information about AI-related savings.
That gets straight to the awkward part of legal AI. A firm can take a research or document-review task that once required several junior-associate hours and finish it much faster. Under a strict hourly model, efficiency reduces billable time. Under a fixed-fee model, efficiency can increase the firm’s margin.
Smaller firms are running into the same issue. Clio found 86% of solo firms and 78% of small firms had made no pricing changes after adopting AI. Only around a third had grown revenue from AI, even though many reported faster work and higher quality.
Pricing is likely to become a much bigger part of the adoption story from here. Clients can tolerate vague AI efficiency claims while the technology is experimental. Once firms are using AI every day for research, drafting and review, clients have a much simpler question: why are they still paying for the old number of hours?
This chart, featured in our Legal Tech market deck, breaks down revenue across customer segments in the Legal Tech market
So what is actually getting real adoption in legal AI now?
Legal AI is getting real adoption today in research, document analysis, contract work, drafting and due diligence, with supervised multi-step agents beginning to join that group.
Legal research has the widest evidence base. Document review and summarization sit very close behind. Drafting has crossed into normal use when the AI produces something a lawyer reviews and edits. Contract work may be the strongest commercial use case because teams can measure review time, turnaround time and outside-counsel costs. Due diligence is already live at major firms because AI can search and structure large sets of agreements much faster than junior lawyers can do manually.
Litigation is moving in the same direction through evidence analysis, chronologies, discovery and drafting. The common thread is that the AI gets a defined job and the lawyer can inspect what came back.
That is also why specialist legal AI is doing so well even though lawyers already have ChatGPT and Claude. Once AI needs access to case law, privileged files, precedents, permissions and firm-specific playbooks, the surrounding system becomes as important as the underlying model.
The newest developments reinforce that direction. Harvey and Legora have reached six-figure professional user bases, Thomson Reuters is pushing CoCounsel deeper into full workflows, Google has entered with Gemini Enterprise for Legal, OpenAI is moving closer to legal software, and firms such as Kirkland and Freshfields are spending heavily on their own AI capabilities.
Agents are the next layer, but the evidence is weaker there. They can already chain together useful legal tasks. Current benchmarks and deployment behavior still point toward a lawyer checking the work before anything consequential happens.
The answer is fairly sharp. Legal AI has already won meaningful pieces of the lawyer’s workflow. Researching the law, reading large document sets, comparing contracts, finding clauses, producing first drafts and preparing diligence are becoming normal AI-assisted work. Giving AI responsibility for the whole matter, the strategy and the final legal judgment remains premature.
The legal AI products getting adopted fastest are the ones that make lawyers dramatically faster while keeping the lawyer close enough to verify the result. That is where real adoption is today.
If you want more recent data on this point, please see our latest Legal Tech market report.
OUR METHODOLOGY
This analysis asks what is actually getting real adoption in legal AI now. We separated occasional experimentation from repeated use on live legal work, then compared evidence across research, document review, drafting, contracting, due diligence, litigation, agent workflows, firm size, in-house teams and legal AI platforms.
We prioritized recent surveys, production-deployment data and first-hand disclosures that could show how lawyers are actually using AI. Repeated usage, organizational deployment, workflow integration and measurable outcomes carried more weight than partnership announcements, product launches or stated interest in adopting AI.
Different surveys use different definitions of AI adoption, so we did not combine their percentages into a single market-wide rate. Instead, we compared the direction of the evidence across different samples and looked for areas where independent sources kept pointing toward the same workflows.
Vendor disclosures were used mainly for facts vendors can directly observe, such as customer counts, active usage, query volume, workflow deployments and product integrations. Claims about productivity or business impact were treated more cautiously when the underlying research came from companies selling the technology being studied.
For AI agents, we distinguished between multi-step workflow automation and autonomous responsibility for an entire legal matter. Production workflow evidence was considered alongside benchmark results because the two answer different questions: whether lawyers are using agentic systems today, and how reliably those systems can complete complex legal work end to end.
Key research sources include Thomson Reuters’ 2026 AI in Professional Services Report, Thomson Reuters’ Future of Professionals legal research, LexisNexis’ 2026 legal AI research, the State Bar of Texas AI in the Practice of Law Survey, the SKILLS Legal AI Use Case Survey, the 2026 Legal Industry Report, Clio’s research on solo and small firms, Deloitte Legal’s AI research, LegalOn and In-House Connect’s State of AI for In-House Legal, LegalOn’s In-House Legal Pulse, and Ironclad’s State of AI in Legal.
For product scale and workflow evidence, we also used Harvey’s platform disclosures, Harvey’s customer data, the Maddocks rollout, Harvey’s Legal Agent Benchmark, Legora’s newsroom, Legora’s $100 million ARR disclosure, Thomson Reuters’ CoCounsel scale disclosure, the latest CoCounsel Legal release, Google Cloud’s Gemini Enterprise for Legal announcement, Freshfields’ Anthropic partnership, and Kirkland & Ellis’ Palantir partnership.
This chart, featured in our Legal Tech market deck, shows how AI contract review platform technology has evolved over time
Related blog posts
- LegalTech: what are startups building now?
- Legal AI: what’s changing now?
- Legal AI: what is actually working now?
- The latest funding news in the Legal Tech market
