Yesterday I attended a press event hosted by Thomson Reuters where they announced an demoed a suite of four GenAI products and initiatives. Only three months after completing the acquisition of Casetext they are launching a suite of tools which are driven by the integration of Casetext and CoCounsel technologies with TRs legacy AI infrastructure.
Standing on the Doorstep of Change I have spent forty years watching legal research evolve, so my focus in this post will be on the Westlaw Precision AI functionality. During yesterday’s demo, I had to remind myself “we are not in Kansas anymore.” My default is to expect the product to be more like the research experience am used to in Westlaw Edge. This reminds me of the early days of Westlaw – it started out being a system which could only search the West headnotes and the key number system. It delivered efficiencies over print research but didn’t at all suggest the completely transformed and unimaginable changes that would come. No one was thinking about legal analytics in 1980, but now they are an essential part of legal strategy and legal research. I feel that we are in that same place with AI Assisted Research on Westlaw Precision, It offers a foundational move toward a new way of performing research and integrating research into workflow. I need to remind myself that there is a future of legal research I can’t yet imagine.
The new products include:
• AI-Assisted Research on Westlaw Precision – Better, faster answers to complex research questions drawn from the industry’s most comprehensive collection of editorially enhanced content.
• A new GenAI assistant connects all Thomson Reuters generative AI products, building on innovation from Casetext
• Thomson Reuters Generative AI Platform – A common development platform to design, build, and deploy GenAI skills with unparalleled speed • New GenAI capabilities for Practical Law – Customers to benefit from AI chat-type interface
• CoCounsel Core – Announcing the commercial offering of CoCounsel skills as part of the Thomson Reuters portfolio
“Thomson Reuters is redefining the way legal work is done by delivering a generative AI-based toolkit to enable attorneys to quickly gather deeper insights and deliver a better work product. AI-Assisted Research on Westlaw Precision and CoCounsel Core provide the most comprehensive set of generative AI skills that attorneys can use across their research and workflow,” said David Wong, chief product officer.
AI Assisted Research on Westlaw Precision is now available to Precision customers in the United States. It is described as offering a “best of” approach which combines the technology from both Casetext and the Thomson Reuters Generative AI Platforms. .AI-Assisted Research allows customers to ask complex legal research questions in natural language and receive synthesized answers, with links to supporting authority from Westlaw content and links to further examine that authority. During a pre-launch event Mike Dahn, Head of Westlaw Product pointed out that although AI is an important component of the new product, the technology is rooted in and enhanced by “more than 150 years of Thomson Reuters classification, analysis, and editorial expertise contributed by subject matter experts and attorney editors”.
A few observations:
- AI Assisted Research offers a powerful new approach to researching the law. It allows lawyers to enter natural language queries but does not require any specific type of “prompts.” It is recommended that questions should be a statement which includes all the information a partner might give to an associate when assigning them a research project.
- Cautioning Lawyers The initial screen of AI assistant includes cautionary language. Frankly I with it were more prominent – I suggest outlining the text in a red boarded box or having a “splash screen” with the cautionary language which users have to acknowledge. Here is the text: “Assisted research uses large language models and can occasionally produce inaccuracy, so it should always be used as part of a research process in connection with additional research to fully understand the nuance of the issues and further improve accuracy.”
- Minimizing Risk. Thomson Reuters is using Retrieval Augmented Generation (RAG) to prevent the large language models (LLMs) from making up things like case names and citations by focusing the LLMs on the actual language of Westlaw content.
- How it works. You ask the assistant a question and it delivers an answer which is linked to the underlying precedent. The answer takes longer to deliver than the near-instantaneous search time on Westlaw Edge. It can take up to two minutes to get an answer and if you are impatient like me, there is a “email me when the response is ready” button. I have to admit -that gave me pause, it felt like a step in the wrong direction. I asked Dahn if they had gotten complaints from lawyers regarding the response time. He indicated that they had not, and stated that lawyers who tried the system acknowledged that the system saved time in the overall research process. He also pointed out that that the answer isn’t simply a list of cases – the results include a summary and analysis of the caselaw or other primary law. The bottom line is that they are working on improving the speed.
- Content and functionality – The system can search all 70 US jurisdictions all at once. However there is a limit on how many individual jurisdictions can be combined in a search. A researcher can choose up to THREE individual states plus related federal jurisdictions. It currently lacks secondary sources. I miss the sidebar of related content. There are footnotes instead of “in line” citations. (I believe this will be addressed shortly).
- Conversational Search – up to 8 additional questions can be asked to refine results.
- Training – While the product is fairly intuitive, it is different enough from the Westlaw Edge experience that I think they should have some quick embedded videos to help lawyers get up to speed on how to use the product. There is a “tips for best results link” at the top of the screen. I suspect that firms rolling out this product will want to offer some training classes before they make the product available to new associates.
CoCounsel Core will offer eight Generative AI enabled core skills. including AI-Assisted Research on Westlaw Precision, Prepare for a Deposition, Draft Correspondence, Search a Database, Review Documents, Summarize a Document, Extract Contract Data, and Contract Policy Compliance. CoCounsel Core will remain a commercial offering of legacy CoCounsel skills. CoCounsel is built on more than 4,000 hours of testing and training by more than 1,400 customers since its launch in May.
AI Assistant TR announced plans to launch “world’s first AI legal assistant” in 2024. The product will marketed under the CoCounsel brand and will be built on the CoCounsel technology. The AI Assistant will be fully integrated with multiple Thomson Reuters legal products, including Westlaw Precision, Practical Law Dynamic Tool Set, Document Intelligence, and HighQ, and will continue to be available on the CoCounsel application as a destination site.
Practical Law TR will be adding AI chat assisted research capabilities to Practical Law.
What’s Coming A roadmap for the future was laid out in the press release as follows: Future plans include expanding GenAI throughout the research process in Westlaw and bringing these capabilities to versions of Westlaw outside the US. Starting in 2024, target markets include the UK, Canada, Australia, and New Zealand. “Thomson Reuters is well positioned to deliver high-quality AI results because it has the largest, most upto-date, and trustworthy legal research system in the world,” said Andrew Bedigian, counsel, Larson LLP. “The fact that AI-Assisted Research relies exclusively on Thomson Reuters vetted database should provide lawyers with confidence that the answer AI-Assisted Research is generating in response to attorney questions is going to be well supported. And the fact that AI-Assisted Research delivers all the sources it relied upon – right beneath the answer – provides additional confidence that the program is delivering on our research needs.”
“We leveraged our experts in Thomson Reuters Labs, our more than 1,600 attorneys, and our best-inclass content to build a Westlaw Precision tool that provides our customers with the trust, accuracy, and speed they need to serve their clients,” said Mike Dahn, head of Westlaw Product Management, Thomson Reuters. “Our human oversight, technology expertise, and industry-leading content are critical to producing trusted answers with generative AI.
This tool won’t obviate the need for attorneys, but it will help them do their work better and faster.” CoCounsel Core With integration well under way, the company today announced CoCounsel Core as the commercial offering of legacy CoCounsel skills. CoCounsel has had more than 4,000 hours of rigorous testing and training by more than 1,400 customers since its launch in May – including input of law firms, corporations, and government agencies. All of this learning is being utilized in the Thomson Reuters innovation journey.