{"id":159334,"date":"2026-08-03T08:40:16","date_gmt":"2026-08-03T16:40:16","guid":{"rendered":"https:\/\/xira.com\/p\/2026\/08\/03\/as-ai-legal-research-tools-race-to-give-answers-this-new-product-gives-only-cases-and-thats-the-point\/"},"modified":"2026-08-03T08:40:16","modified_gmt":"2026-08-03T16:40:16","slug":"as-ai-legal-research-tools-race-to-give-answers-this-new-product-gives-only-cases-and-thats-the-point","status":"publish","type":"post","link":"https:\/\/xira.com\/p\/2026\/08\/03\/as-ai-legal-research-tools-race-to-give-answers-this-new-product-gives-only-cases-and-thats-the-point\/","title":{"rendered":"As AI Legal Research Tools Race to Give Answers, This New Product Gives Only Cases \u2014 And That\u2019s the Point"},"content":{"rendered":"

At a time when virtually every major AI legal research product is moving toward delivering a single, synthesized answer to your research question, a new product launching today is based on the premise that many litigators want something that sounds almost old school \u2013 just the cases. Align Research, released to general availability this morning [\u2026]<\/p>\n

At a time when virtually every major AI legal research product is moving toward delivering a single, synthesized answer to your research question, a new product launching today is based on the premise that many litigators want something that sounds almost old school \u2013 just the cases.<\/p>\n

Align Research<\/a>, released to general availability this morning by the litigation technology company Align<\/a>, takes a research question in plain language, goes off for a few hours, and comes back with a set of relevant court opinions, with the key passages highlighted, and nothing else.<\/p>\n

There is no memo, no summary, no answer. As the company\u2019s press release puts it, the product \u201cretrieves rather than generates.\u201d<\/p>\n

Align founder and CEO Sam Davidoff<\/a>, a former litigation partner at Williams & Connolly, said the product\u2019s design is in part an answer to the hallucination problem that has plagued the use of generative AI in law, and in part a return to what he believes is the actual task litigators most often want done.<\/p>\n

\u201cIt\u2019s not telling you anything, it\u2019s just showing you,\u201d Davidoff told me in an interview about the launch on my LawNext PR podcast<\/a>. \u201cIf all the AI is doing is essentially pointing you to pre-existing things, as opposed to writing some narrative for you, it actually can\u2019t hallucinate. It could miss something, it could highlight something that you don\u2019t agree with, but it can\u2019t make something up.\u201d<\/p>\n

\u2018Get Me the Cases\u2019<\/strong><\/h3>\n

Davidoff said he had long resisted entering the legal research market, which he considered well saturated. What changed his mind was a gap he perceived in the current crop of AI research tools, rooted in his own years of practice.<\/p>\n

The research task he most often performed as an associate, and most often assigned as a partner, was not, \u201cAnswer this question,\u201d he said, but, \u201cGet me the cases. Find me every district court that has ever ruled this way on a motion to dismiss.\u201d<\/p>\n

Most AI legal research products these days are built around the opposite model, where you ask a question and get back a researched and validated answer. \u201cThat\u2019s a little different than what, at least a lot of times, litigators want,\u201d Davidoff said.<\/p>\n

When I observed that this sounds, in a way, old school \u2013 closer to how we used to do research \u2013 he readily agreed. \u201cThere\u2019s no secret where I get my ideas from. I was practicing law and this is how we used to do it, and there\u2019s a little part of me that says, \u2018Hey, some of our old tricks still kind of work.\u2019\u201d<\/p>\n

In a LinkedIn post<\/a> ahead of today\u2019s launch, Davidoff listed some of the ideas that went into the product relating to the proper division of labor between lawyers and AI. Among them are, \u201cDon\u2019t delegate what you should learn,\u201d and, \u201cAI should show, not tell.\u201d<\/p>\n

Give the AI the uninteresting work of churning through thousands of cases, he said, and keep the interesting work of reading and analyzing the few dozen that matter for the lawyer to do.<\/p>\n

\u201cIf you give to AI the job of actually reading the cases and distilling the final analysis, it means you\u2019re not doing it,\u201d he told me. \u201cThat\u2019s both sometimes the fun of being a lawyer, but also sometimes the strategic advantage of being a lawyer \u2013 that I\u2019ve read the cases and I see a new angle for how we could argue this in front of this judge.\u201d<\/p>\n

How It Works<\/strong><\/h3>\n

As Davidoff describes it, Align Research is an agentic pipeline. One model analyzes the question and breaks out the legal research issues. Another decides what searches to run. Others read the retrieved cases, classify them, make relevance decisions, and determine whether more searching is needed. A final stage decides which passages to highlight.<\/p>\n