{"id":111851,"date":"2025-03-25T15:59:36","date_gmt":"2025-03-25T23:59:36","guid":{"rendered":"https:\/\/xira.com\/p\/2025\/03\/25\/deep-research\/"},"modified":"2025-03-25T15:59:36","modified_gmt":"2025-03-25T23:59:36","slug":"deep-research","status":"publish","type":"post","link":"https:\/\/xira.com\/p\/2025\/03\/25\/deep-research\/","title":{"rendered":"Deep Research"},"content":{"rendered":"
\"\"
<\/figcaption><\/figure>\n

One of the greatest challenges of Generative AI solutions like ChatGPT is hallucination. They create fear in professionals as well as nightmares involving \u201cthe lawyer that used ChatGPT\u201d and filed a fictitious case with a judge.<\/p>\n

Hallucinations are a problem because GenAI solutions like ChatGPT are not databases, even though they appear to behave like search engines that provide a custom answer to a question.\u00a0The reality is Large Language Models (LLMs) are spitting out probabilistic answers one character at a time. They can be incredibly convincing and accurate, but they can also be incredibly convincing and hallucinate too.<\/p>\n

The lines between GenAI and search engines have been blurring though.\u00a0Retrieval Augmented Generation (RAG) is a technique that connects the power of GenAI with trusted data pulled from reliable sources. This is how many commercial offerings including some legal research services incorporate GenAI. They search and retrieve trusted information in a database and then limit the scope of how the LLM is used. More basic capabilities like summarizing relevant portions of the trusted material to provide an answer are common. Asking an LLM to summarize specific documents bounds the probabilistic output to the content within the documents selected.\u00a0 The result is a much more reliable answer that is less prone to hallucinations or inaccuracies. It also explains how systems can provide links and citations to the underlying material.<\/p>\n

RAG blends reliable content sets with GenAI to create more trustworthy systems where hallucinations and errors are far less likely to occur, while allowing users or other professionals to check and validate the work.<\/p>\n

Enter Deep Research<\/strong><\/p>\n

So what if ChatGPT or another GenAI solution could search the internet as part of answering a question? What if that offering could also provide links and references to those documents too? And what if that offering could use chain of thought reasoning to perform multiple searches and then explain the steps it took to answer your question?<\/p>\n

Well that explains OpenAI\u2019s Deep Research service that was announced earlier this year.\u00a0OpenAI has added a RAG capability against Internet data and has also incorporated Agentic technology to break down a complex research question into multiple searches in support of discrete tasks.<\/p>\n

Within seven days of the release of ChatGPT Turbo, I heard about a student who submitted a paper using ChatGPT. (The paper got an \u201cA\u201d by the way.) Now a student looking for an outline for a college research paper can have a draft of the paper with internet-sourced citations in a matter of minutes<\/a>. The academic community is already wrestling with the issues. How will the academic community adjust? Similarly, many professionals may find Deep Research to be a friend or an eventual enemy<\/a>.<\/p>\n

Implications For Legal Professionals<\/strong><\/p>\n

Let me suggest that it\u2019s just a matter of time before Microsoft Copilot will have Deep Research integration. There\u2019s a potential future in which Copilot will be able to generate answers to research questions that will include information from the public internet and from internal data sources and perhaps third-party vendor databases.\u00a0(Recall that OpenAI has a relationship with Microsoft.) If that happens, expect competitors like Google\u2019s Gemini and\u00a0 Anthropic\u2019s Claude to follow suit with similar offerings.<\/p>\n

While the academic community and professionals like marketers may find tremendous value and disruption, what do products like Deep Research mean to the legal profession?<\/p>\n

Clients Will Expect More<\/strong><\/p>\n

Whether we are talking about a law department\u2019s internal clients or a law firm\u2019s external clients, there will be higher expectations for advice.\u00a0Why?\u00a0Because a client\u2019s ability to ask a research question and get a preliminary answer just got better. The bar has been raised by increasing the practical knowledge and use of logic within GenAI solutions.\u00a0Even if the legal logic and legal knowledge is limited or flawed, clients will be better equipped before they ever reach out to an attorney. <\/p>\n

Deep Research Coupled With Proprietary Data Empowers Firms<\/strong><\/p>\n

If my prediction is true and attorneys may one day be able to use Copilot to generate answers that include referenceable information from the internet and their internal data, templates, and client data, the time to create a first draft of work product will be reduced dramatically. There will still be review required and some augmentation with traditional research and human input, but the process will be accelerated.<\/p>\n

If legal research vendors make their data available in the same Copilot application, the nirvana envisioned by those that have pushed for federated search solutions will be achieved tenfold.<\/p>\n

But What About Legal Research Providers?<\/strong><\/p>\n

On the surface one might assume that Copilot paired with a Deep Research capability, access to internal data, vendor data, and the internet would be the only legal research solution necessary for a law firm. \u00a0\u00a0<\/p>\n

The reality is we will be multimodal for the foreseeable future.\u00a0Recall there are still attorneys that prefer to work in paper and in print.\u00a0There are Boolean searchers.\u00a0And there are different generations of research platforms, including many without AI features.\u00a0These aren\u2019t going away anytime soon. If Deep Research-style functionality is the next new horizon, there are multiple past horizons still awaiting sunset.<\/p>\n

There will also be a significant role for deeper vertical solutions within legal and other industries. The internet did not put legal research providers out of business, nor did Wikipedia.\u00a0There was virtually no impact.\u00a0I think a similar outcome is likely with Deep Research-style functionality.\u00a0Why?\u00a0Because companies like Microsoft, Google, Apple, Oracle, or SAP operate at scale. They can\u2019t handle the corner cases and nuances of specialists in a vast number of industries.\u00a0That would keep them from operating at scale. Add to that regulations and the nuances of legal advice and it\u2019s entirely likely that legal-specific platforms will embrace Deep Research-style functionality too.\u00a0OpenAI doesn\u2019t have dedicated account management and support lines, let alone experts within the legal industry.<\/p>\n

Legal professionals and professionals in other mission critical industries will find value in Deep Research, just as many attorneys start their research on the internet.<\/p>\n

But, while there will be value in those broad-based research services, there will continue to be a role and demand for deep industry vertical players that provide comprehensive service and deeper industry specific functionality and data.<\/p>\n

\n

Ken Crutchfield is Vice President and General Manager of Legal Markets at Wolters Kluwer Legal & Regulatory U.S., a leading provider of information, business intelligence, regulatory and legal workflow solutions. Ken has more than three decades of experience as a leader in information and software solutions across industries. He can be reached at\u00a0<\/strong><\/em>ken.crutchfield@wolterskluwer.com<\/em><\/strong><\/a>.<\/strong><\/em><\/p>\n

The post Deep Research<\/a> appeared first on Above the Law<\/a>.<\/p>\n

\"\"
<\/figcaption><\/figure>\n

One of the greatest challenges of Generative AI solutions like ChatGPT is hallucination. They create fear in professionals as well as nightmares involving \u201cthe lawyer that used ChatGPT\u201d and filed a fictitious case with a judge.<\/p>\n

Hallucinations are a problem because GenAI solutions like ChatGPT are not databases, even though they appear to behave like search engines that provide a custom answer to a question.\u00a0The reality is Large Language Models (LLMs) are spitting out probabilistic answers one character at a time. They can be incredibly convincing and accurate, but they can also be incredibly convincing and hallucinate too.<\/p>\n

The lines between GenAI and search engines have been blurring though.\u00a0Retrieval Augmented Generation (RAG) is a technique that connects the power of GenAI with trusted data pulled from reliable sources. This is how many commercial offerings including some legal research services incorporate GenAI. They search and retrieve trusted information in a database and then limit the scope of how the LLM is used. More basic capabilities like summarizing relevant portions of the trusted material to provide an answer are common. Asking an LLM to summarize specific documents bounds the probabilistic output to the content within the documents selected.\u00a0 The result is a much more reliable answer that is less prone to hallucinations or inaccuracies. It also explains how systems can provide links and citations to the underlying material.<\/p>\n

RAG blends reliable content sets with GenAI to create more trustworthy systems where hallucinations and errors are far less likely to occur, while allowing users or other professionals to check and validate the work.<\/p>\n

Enter Deep Research<\/strong><\/p>\n

So what if ChatGPT or another GenAI solution could search the internet as part of answering a question? What if that offering could also provide links and references to those documents too? And what if that offering could use chain of thought reasoning to perform multiple searches and then explain the steps it took to answer your question?<\/p>\n

Well that explains OpenAI\u2019s Deep Research service that was announced earlier this year.\u00a0OpenAI has added a RAG capability against Internet data and has also incorporated Agentic technology to break down a complex research question into multiple searches in support of discrete tasks.<\/p>\n

Within seven days of the release of ChatGPT Turbo, I heard about a student who submitted a paper using ChatGPT. (The paper got an \u201cA\u201d by the way.) Now a student looking for an outline for a college research paper can have a draft of the paper with internet-sourced citations in a matter of minutes<\/a>. The academic community is already wrestling with the issues. How will the academic community adjust? Similarly, many professionals may find Deep Research to be a friend or an eventual enemy<\/a>.<\/p>\n

Implications For Legal Professionals<\/strong><\/p>\n

Let me suggest that it\u2019s just a matter of time before Microsoft Copilot will have Deep Research integration. There\u2019s a potential future in which Copilot will be able to generate answers to research questions that will include information from the public internet and from internal data sources and perhaps third-party vendor databases.\u00a0(Recall that OpenAI has a relationship with Microsoft.) If that happens, expect competitors like Google\u2019s Gemini and\u00a0 Anthropic\u2019s Claude to follow suit with similar offerings.<\/p>\n

While the academic community and professionals like marketers may find tremendous value and disruption, what do products like Deep Research mean to the legal profession?<\/p>\n

Clients Will Expect More<\/strong><\/p>\n

Whether we are talking about a law department\u2019s internal clients or a law firm\u2019s external clients, there will be higher expectations for advice.\u00a0Why?\u00a0Because a client\u2019s ability to ask a research question and get a preliminary answer just got better. The bar has been raised by increasing the practical knowledge and use of logic within GenAI solutions.\u00a0Even if the legal logic and legal knowledge is limited or flawed, clients will be better equipped before they ever reach out to an attorney. <\/p>\n

Deep Research Coupled With Proprietary Data Empowers Firms<\/strong><\/p>\n

If my prediction is true and attorneys may one day be able to use Copilot to generate answers that include referenceable information from the internet and their internal data, templates, and client data, the time to create a first draft of work product will be reduced dramatically. There will still be review required and some augmentation with traditional research and human input, but the process will be accelerated.<\/p>\n

If legal research vendors make their data available in the same Copilot application, the nirvana envisioned by those that have pushed for federated search solutions will be achieved tenfold.<\/p>\n

But What About Legal Research Providers?<\/strong><\/p>\n

On the surface one might assume that Copilot paired with a Deep Research capability, access to internal data, vendor data, and the internet would be the only legal research solution necessary for a law firm. \u00a0\u00a0<\/p>\n

The reality is we will be multimodal for the foreseeable future.\u00a0Recall there are still attorneys that prefer to work in paper and in print.\u00a0There are Boolean searchers.\u00a0And there are different generations of research platforms, including many without AI features.\u00a0These aren\u2019t going away anytime soon. If Deep Research-style functionality is the next new horizon, there are multiple past horizons still awaiting sunset.<\/p>\n

There will also be a significant role for deeper vertical solutions within legal and other industries. The internet did not put legal research providers out of business, nor did Wikipedia.\u00a0There was virtually no impact.\u00a0I think a similar outcome is likely with Deep Research-style functionality.\u00a0Why?\u00a0Because companies like Microsoft, Google, Apple, Oracle, or SAP operate at scale. They can\u2019t handle the corner cases and nuances of specialists in a vast number of industries.\u00a0That would keep them from operating at scale. Add to that regulations and the nuances of legal advice and it\u2019s entirely likely that legal-specific platforms will embrace Deep Research-style functionality too.\u00a0OpenAI doesn\u2019t have dedicated account management and support lines, let alone experts within the legal industry.<\/p>\n

Legal professionals and professionals in other mission critical industries will find value in Deep Research, just as many attorneys start their research on the internet.<\/p>\n

But, while there will be value in those broad-based research services, there will continue to be a role and demand for deep industry vertical players that provide comprehensive service and deeper industry specific functionality and data.<\/p>\n

\n

Ken Crutchfield is Vice President and General Manager of Legal Markets at Wolters Kluwer Legal & Regulatory U.S., a leading provider of information, business intelligence, regulatory and legal workflow solutions. Ken has more than three decades of experience as a leader in information and software solutions across industries. He can be reached at\u00a0<\/strong><\/em>[email\u00a0protected]<\/em><\/strong><\/a>.<\/strong><\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

One of the greatest challenges of Generative AI solutions like ChatGPT is hallucination. They create fear in professionals as well as nightmares involving \u201cthe lawyer that used ChatGPT\u201d and filed a fictitious case with a judge. Hallucinations are a problem because GenAI solutions like ChatGPT are not databases, even though they appear to behave like […]<\/p>\n","protected":false},"author":3,"featured_media":111852,"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-111851","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\/03\/artificial-intelligence-3382507_1280-qFfvyE.jpeg?fit=1280%2C853&ssl=1","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/posts\/111851","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=111851"}],"version-history":[{"count":0,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/posts\/111851\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/media\/111852"}],"wp:attachment":[{"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/media?parent=111851"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/categories?post=111851"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/xira.com\/p\/wp-json\/wp\/v2\/tags?post=111851"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}