{"id":136376,"date":"2025-11-05T22:10:06","date_gmt":"2025-11-06T06:10:06","guid":{"rendered":"https:\/\/xira.com\/p\/2025\/11\/05\/the-elephant-in-the-room-for-legal-ai-remains-elephantine-document-sets\/"},"modified":"2025-11-05T22:10:06","modified_gmt":"2025-11-06T06:10:06","slug":"the-elephant-in-the-room-for-legal-ai-remains-elephantine-document-sets","status":"publish","type":"post","link":"https:\/\/xira.com\/p\/2025\/11\/05\/the-elephant-in-the-room-for-legal-ai-remains-elephantine-document-sets\/","title":{"rendered":"The Elephant In The Room For Legal AI Remains Elephantine Document Sets"},"content":{"rendered":"

\u201cThe dirty secret,\u201d a veteran eDiscovery consultant told me, \u201cis that AI still can\u2019t actually handle these massive document sets.\u201d While the industry rushes toward AI adoption, embracing its demonstrated strengths and \u2014 for some insane and persistent reason \u2014 its well-documented weaknesses<\/a>, one hurdle it can\u2019t quite surmount is the ever-expanding size of discovery. AI is, rightly, lauded for its capacity to analyze and summarize large amounts of data, but crunching 100 deposition transcripts is one thing and getting on top of terabytes worth of discovery is quite another. Context windows just aren\u2019t that big.<\/p>\n

Which means legal tech vendors have to get creative within the limits of the technology. Today, at its SYNERGY Legal Professionals conference<\/a>, Thomson Reuters announced a new beta feature that gets users closer.<\/p>\n

CoCounsel Legal\u2019s new bulk document review feature allows analysis of 10,000 documents at a clip, returning sortable, user-friendly results. The secret sauce is in taking these big batches and returning structured analysis that users can then \u2014 for lack of a better term \u2014 daisy chain to get on top of \u201chundreds or thousands of documents far more efficiently than traditional manual methods,\u201d as the press release puts it. <\/p>\n

This underscores, again, that AI isn\u2019t replacing humans<\/a>. Someone has to run the requisite 10K at a time searches and deal with the results. But it will reduce<\/em> the number of humans necessary to get through these traditionally brute force searches and then put everything together. The general purpose AI companies will keep promising exponential growth, but it\u2019s just not coming<\/a> and the key differentiator for legal tech will be building the best workarounds to deal with AI\u2019s baked in context window limitations.<\/p>\n

Thomson Reuters also announced an agentic feature for \u201cIndependent Execution of Legal Tasks.\u201d Sigh. The record will reflect<\/a> my ongoing disdain for the term \u201cagentic.\u201d Independent research into these so-called \u201cagentic\u201d AI tools reveals a dismal track record, with failure rates upward of 70 percent according to a Carnegie Mellon study<\/a>. On top of that, the term is employed by the industry to describe an AI tool that autonomously creates a multi-step workflow and then acts on it, which is just an invitation to multiple points of failure that \u2014 even if it worked better than the studies indicate \u2014 should terrify a lawyer. On top of that, compressing multiple steps in the lawyerly process invites a psychological reordering that can make lawyers, well\u2026 dumber<\/a>.<\/p>\n

But while the new Thomson Reuters beta tool will experiment with independent work, the planned end state<\/em> of the feature is a very useful tool that shouldn\u2019t worry lawyers at all:<\/p>\n

\n

With the second phase, legal professionals will have access to a customizable workflow builder that will allow them to create, save and share their own workflows within CoCounsel Legal capitalizing on the trusted content and solutions from Thomson Reuters as well as the firm\u2019s knowledge. Additionally, this facilitates the development of repeatable, law firm-specific processes and supports reusability across teams \u2013 ensuring consistency while capturing institutional knowledge and best practices.<\/p>\n<\/blockquote>\n

Building a workflow based on the lived experience of the lawyers is exactly the sort of tool a firm can use. It\u2019s \u201cautonomous\u201d to the extent it checks and acts upon the work it\u2019s trained by the firm to accomplish. As with the document review feature, the race here is to build something that allows AI to be better harnessed to match what lawyers actually do.<\/p>\n

If the AI bubble doesn\u2019t explode more epically than that infamous Subway sandwich<\/a>, this is where the action will be for the next few years. Don\u2019t watch the AI as much as what companies like Thomson Reuters are putting on top of AI. That\u2019s the fun stuff and the real differentiator.<\/p>\n

\n

\"Headshot\"Joe Patrice<\/a>\u00a0is a senior editor at Above the Law and co-host of Thinking Like A Lawyer<\/a>. Feel free to\u00a0email<\/a> any tips, questions, or comments. Follow him on\u00a0Twitter<\/a>\u00a0or Bluesky<\/a> if you\u2019re interested in law, politics, and a healthy dose of college sports news. Joe also serves as a Managing Director at RPN Executive Search<\/a>.<\/em><\/strong><\/p>\n

The post The Elephant In The Room For Legal AI Remains Elephantine Document Sets<\/a> appeared first on Above the Law<\/a>.<\/p>\n

\"\"<\/figure>\n

\u201cThe dirty secret,\u201d a veteran eDiscovery consultant told me, \u201cis that AI still can\u2019t actually handle these massive document sets.\u201d While the industry rushes toward AI adoption, embracing its demonstrated strengths and \u2014 for some insane and persistent reason \u2014 its well-documented weaknesses<\/a>, one hurdle it can\u2019t quite surmount is the ever-expanding size of discovery. AI is, rightly, lauded for its capacity to analyze and summarize large amounts of data, but crunching 100 deposition transcripts is one thing and getting on top of terabytes worth of discovery is quite another. Context windows just aren\u2019t that big.<\/p>\n

Which means legal tech vendors have to get creative within the limits of the technology. Today, at its SYNERGY Legal Professionals conference<\/a>, Thomson Reuters announced a new beta feature that gets users closer.<\/p>\n

CoCounsel Legal\u2019s new bulk document review feature allows analysis of 10,000 documents at a clip, returning sortable, user-friendly results. The secret sauce is in taking these big batches and returning structured analysis that users can then \u2014 for lack of a better term \u2014 daisy chain to get on top of \u201chundreds or thousands of documents far more efficiently than traditional manual methods,\u201d as the press release puts it. <\/p>\n

This underscores, again, that AI isn\u2019t replacing humans<\/a>. Someone has to run the requisite 10K at a time searches and deal with the results. But it will reduce<\/em> the number of humans necessary to get through these traditionally brute force searches and then put everything together. The general purpose AI companies will keep promising exponential growth, but it\u2019s just not coming<\/a> and the key differentiator for legal tech will be building the best workarounds to deal with AI\u2019s baked in context window limitations.<\/p>\n

Thomson Reuters also announced an agentic feature for \u201cIndependent Execution of Legal Tasks.\u201d Sigh. The record will reflect<\/a> my ongoing disdain for the term \u201cagentic.\u201d Independent research into these so-called \u201cagentic\u201d AI tools reveals a dismal track record, with failure rates upward of 70 percent according to a Carnegie Mellon study<\/a>. On top of that, the term is employed by the industry to describe an AI tool that autonomously creates a multi-step workflow and then acts on it, which is just an invitation to multiple points of failure that \u2014 even if it worked better than the studies indicate \u2014 should terrify a lawyer. On top of that, compressing multiple steps in the lawyerly process invites a psychological reordering that can make lawyers, well\u2026 dumber<\/a>.<\/p>\n

But while the new Thomson Reuters beta tool will experiment with independent work, the planned end state<\/em> of the feature is a very useful tool that shouldn\u2019t worry lawyers at all:<\/p>\n

\n

With the second phase, legal professionals will have access to a customizable workflow builder that will allow them to create, save and share their own workflows within CoCounsel Legal capitalizing on the trusted content and solutions from Thomson Reuters as well as the firm\u2019s knowledge. Additionally, this facilitates the development of repeatable, law firm-specific processes and supports reusability across teams \u2013 ensuring consistency while capturing institutional knowledge and best practices.<\/p>\n<\/blockquote>\n

Building a workflow based on the lived experience of the lawyers is exactly the sort of tool a firm can use. It\u2019s \u201cautonomous\u201d to the extent it checks and acts upon the work it\u2019s trained by the firm to accomplish. As with the document review feature, the race here is to build something that allows AI to be better harnessed to match what lawyers actually do.<\/p>\n

If the AI bubble doesn\u2019t explode more epically than that infamous Subway sandwich<\/a>, this is where the action will be for the next few years. Don\u2019t watch the AI as much as what companies like Thomson Reuters are putting on top of AI. That\u2019s the fun stuff and the real differentiator.<\/p>\n

\n

\"Headshot\"Joe Patrice<\/a>\u00a0is a senior editor at Above the Law and co-host of Thinking Like A Lawyer<\/a>. Feel free to\u00a0email<\/a> any tips, questions, or comments. Follow him on\u00a0Twitter<\/a>\u00a0or Bluesky<\/a> if you\u2019re interested in law, politics, and a healthy dose of college sports news. Joe also serves as a Managing Director at RPN Executive Search<\/a>.<\/em><\/strong><\/p>\n","protected":false},"excerpt":{"rendered":"

\u201cThe dirty secret,\u201d a veteran eDiscovery consultant told me, \u201cis that AI still can\u2019t actually handle these massive document sets.\u201d While the industry rushes toward AI adoption, embracing its demonstrated strengths and \u2014 for some insane and persistent reason \u2014 its well-documented weaknesses, one hurdle it can\u2019t quite surmount is the ever-expanding size of discovery. […]<\/p>\n","protected":false},"author":3,"featured_media":136373,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_et_pb_use_builder":"","_et_pb_old_content":"","_et_gb_content_width":"","_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[16],"tags":[],"class_list":["post-136376","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-above_the_law"],"jetpack_featured_media_url":"https:\/\/i0.wp.com\/xira.com\/p\/wp-content\/uploads\/2025\/11\/Headshot-300x200-urH7bI.jpg?fit=300%2C200&ssl=1","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/posts\/136376","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/comments?post=136376"}],"version-history":[{"count":0,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/posts\/136376\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/media\/136373"}],"wp:attachment":[{"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/media?parent=136376"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/categories?post=136376"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/tags?post=136376"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}