Thomson Reuters announces new AI initiatives and CoCounsel integration
Thomson Reuters announced Wednesday several new generative artificial intelligence products to aid in legal research while promising additional tools in 2024.
The company unveiled a new AI-Assisted Research feature for its legal research tool Westlaw Precision. The AI-Assisted Research product launched Wednesday allows attorneys to type in legal research questions and get answers supported by links from Westlaw content.
A Thomson Reuters news release said the product will speed up legal research and provide relevant caselaw by drawing on the company’s “comprehensive collection of editorially enhanced content.”
On Wednesday, Thomson Reuters also released CoCounsel Core, “the commercial offering of legacy CoCounsel skills” that attorneys can use for legal research. Among the skills or features, attorneys can prepare a deposition, draft correspondence, search a database, review and summarize documents, and extract contract data.
CoCounsel is an AI legal assistant built on AI research company OpenAI’s GPT-4 large language model and streamlines legal research, deposition preparation and document review. It is software that Thomson Reuters acquired in August as part of a $650 million cash buyout of the startup Casetext.
Meanwhile, Thomson Reuters also announced that it will launch its version of CoCounsel in 2024. The AI assistant will be integrated into Westlaw Precision and other Thomson Reuters legal products and function as the main interface for the company’s generative AI tools, according to the Thomson Reuters news release.
At a press briefing Tuesday, Mike Dahn, head of Westlaw product management, said the company thought that the AI-Assisted Research feature was “the biggest transformational change to come to legal research since the move from book research to online research.”
“We’ve all come together to build something that dramatically improves your research,” Dahn said.
Dahn said the product’s makers had worked to “dramatically reduce hallucinations,” an ongoing problem in generative AI technology in which the software can produce citations and references that appear legitimate but the technology has made up.
Thomson Reuters is using an AI framework called “retrieval-augmented generation,” he added, a technique that guards against hallucinations by having the tech draw upon the most up-to-date and reliable information when producing an answer.
According to Dahn, the framework largely curbs the problem of hallucinations. He said the bigger problem is when the technology simply gets an answer wrong.
“Accuracy is what we care about most, and that’s what we’re most focused on,” Dahn said.
In prepared remarks released after the briefing, Dahn said the company was relying on “human oversight, technology expertise and industry-leading content,” so its tech would produce accurate responses.
Meanwhile, on Tuesday, competitors LexisNexis announced software called Lexis Snapshot and Lexis Create. Lexis Snapshot creates summaries of civil complaints in federal courts. Lexis Create is a drafting tool that works in Microsoft Word.