<p>The legal profession is being transformed radically by the emergence of powerful artificial intelligence (AI) models and agentic AI. There are AI assistants, connectors, and embedded models that, as Thomson Reuters’ chief technology officer, Joel Hron, notes in a <a href="https://www.thomsonreuters.com/en-us/posts/innovation/starting-the-work-is-easy-defending-the-work-is-what-matters/">recent blog</a>, make it possible to start work faster and from more places than ever before.</p> <p>He says AI can now draft, summarise, and analyse in seconds, which changes how legal work begins. The hard part is ensuring outputs are grounded in authoritative sources, validated for accuracy, and traceable back to their origin and Thomson Reuters has spent the last three years working on how to deliver an agentic AI workflow for the legal profession.</p> <p>About nine months ago, the company decided to rebuild CoCounsel from the ground up, taking what Hron calls “an agent-first approach with coding agent orientation in partnership with Anthropic. He says: “We focused on giving access to our content tools from the agent natively and we have exposed all our information so that the agent can use these tools to discover, plan and execute work.”</p> <p>Among the areas of the development he is most proud of is an evaluation framework called CoCounsel Bench. He says: “We spent hundreds of thousands of man-hours with our internal experts developing to enable us to climb the hill from an agent development standpoint and help us understand whether we are making progress across all of the different segments of law.” </p> <p>While 99.99% of the development effort has been spent on CoCounsel, there is now support for agentic AI workflows using Model Context Protocol (MCP). MCP support has mainly involved authentication and making sure that the connection is secure and private and that customer data is protected in an appropriate way. </p> <p>Product development began in 2023 with the $650m acquisition of CaseText, which had developed CoCounsel as an AI legal assistant powered by GPT-4. In May, Thomson Reuters announced new MCP integration with Anthropic that connects Claude directly to CoCounsel Legal. This enables legal professionals to move seamlessly between general-purpose AI and citation-grounded legal work, from either working environment as Hron explains: “The way to think about the two Anthropic Claude and CoCounsel is general AI versus industry-specific AI. In a legal context, it’s the business of law versus the practice of law.”</p> <p>Although the MCP connector can be invoked by a Claude user to communicate with CoCounsel, Hron says that it is more likely that Claude will access CoCounsel itself, to find the information it needs to complete a task.</p> <p>“It will come up with its own plan and trajectory of work, cycling through iterations as many times as it deems necessary, which is why we have wide error bars in terms of the expectations of how those interactions with CoCounsel will happen,” he says. Patterns of usage are analysed to enforce rate limining and restrictions to prevent CoCounsel from being overloaded. “These constructs are very similar to how you might think of an API rate limit, and, in many ways they're identical,” he adds.</p>
Agentic workflows: A Computer Weekly Downtime Upload podcast
<p>The legal profession is being transformed radically by the emergence of powerful artificial intelligence (AI) models and agentic AI. There are AI assistants, connectors, and embedded models that, as Thomson Reuters’ chief technology officer, Joel Hron, notes in a <a href="https://www.thomsonreuters.com/en-us/posts/innovation/starting-the-work-is-easy-defending-the-work-is-what-matters/">recent blog</a>, make it possible to start work faster and from more places than ever before.</p> <p>He says AI can now draft, summarise, and analyse in seconds, which changes how legal work begins. The hard part is ensuring outputs are grounded in authoritative sources, validated for accuracy, and traceable back to their origin and Thomson Reuters has spent the last three years working on how to deliver an agentic AI workflow for the legal profession.</p> <p>About nine months ago, the company decided to rebuild CoCounsel from the ground up, taking what Hron calls “an agent-first approach with coding agent orientation in partnership with Anthropic. He says: “We focused on giving access to our content tools from the agent natively and we have exposed all our information so that the agent can use these tools to discover, plan and execute work.”</p> <p>Among the areas of the development he is most proud of is an evaluation framework called CoCounsel Bench. He says: “We spent hundreds of thousands of man-hours with our internal experts developing to enable us to climb the hill from an agent development standpoint and help us understand whether we are making progress across all of the different segments of law.” </p> <p>While 99.99% of the development effort has been spent on CoCounsel, there is now support for agentic AI workflows using Model Context Protocol (MCP). MCP support has mainly involved authentication and making sure that the connection is secure and private and that customer data is protected in an appropriate way. </p> <p>Product development began in 2023 with the $650m acquisition of CaseText, which had developed CoCounsel as an AI legal assistant powered by GPT-4. In May, Thomson Reuters announced new MCP integration with Anthropic that connects Claude directly to CoCounsel Legal. This enables legal professionals to move seamlessly between general-purpose AI and citation-grounded legal work, from either working environment as Hron explains: “The way to think about the two Anthropic Claude and CoCounsel is general AI versus industry-specific AI. In a legal context, it’s the business of law versus the practice of law.”</p> <p>Although the MCP connector can be invoked by a Claude user to communicate with CoCounsel, Hron says that it is more likely that Claude will access CoCounsel itself, to find the information it needs to complete a task.</p> <p>“It will come up with its own plan and trajectory of work, cycling through iterations as many times as it deems necessary, which is why we have wide error bars in terms of the expectations of how those interactions with CoCounsel will happen,” he says. Patterns of usage are analysed to enforce rate limining and restrictions to prevent CoCounsel from being overloaded. “These constructs are very similar to how you might think of an API rate limit, and, in many ways they're identical,” he adds.</p>
