Today Thomson Reuters announced the launch of CoCounsel Legal which features both Deep Research and agentic workflows.  I attended a recent press briefing where Raghu Ramanathan, president of Legal Professionals, Laura Clayton McDonnnell, President of Corporate Business Rawia Ashraf,  Head of Product, CoCounsel Transactional and GCO, and Mike Dahn Head of Product Management Westlaw demonstrated a raft of new capabilities that resets the bar for AI enabled legal workflows.

Since CoCounsel has evolved several times of the past few years I ask TR to provide some background on the scope and market position of the product. Here is the response:

 To keep pace with the rapid evolution of AI, we’re introducing a revised commercial model that ensures clients benefit from cutting-edge advancements as they emerge. CoCounsel Legal will be offered through multi-year subscriptions with modestly increased annual fees. These subscriptions include continual upgrade releases at no additional cost, delivering ongoing innovation and value. Here’s how the structure now works:

Deep Research (or Westlaw Talks Back)

For several years I have been musing of a future in which research products will provide conversational suggestions not just prompts. Deep Research provides an actual research plan for the attorney. Today Deep Research is part of the latest and according to Dahn the last version of Westlaw and is branded as Westlaw Advantage.

Deep Research can:

• Generate multi-step research plans

• Trace its logic with transparent reasoning

• Deliver structured, Westlaw and Practical Law citation-backed reports

I will be writing a separate post to explore the many new features of Westlaw Advantage.

Guided Workflows

Several CoCounsel guided workflows were launched in July. Soon-to-be-launched workflows include:

 • Draft a Privacy Policy

• Draft an Employee Policy

• Draft a complaint

• Draft discovery request

• Draft discovery response

• Deposition transcript review

Agentic AI vs Assistive AI

 TR provided the following information on how they define and distinguish agentic and assistive AI. Here is the response:

  • An AI assistant responds to prompts and performs single-step tasks, requiring user input at each stage.
  • An AI agent can plan, reason, and act independently within a workflow. It does not just respond, it executes.

AI Agents: First They use advanced reasoning models, supported by an AI assistant that can help orchestrate work. Second, they have access to tools, and they can use these tools to complete tasks.  Third, they can adapt and respond to new information, changing course as needed to achieve their outcomes.  Fourth, as a result of these capabilities, agentic AI systems can complete complex, multi-step assignments.

Knowledge Search – Uncovering and exploiting law firm knowledge

Embedded Training

One of the key unarticulated benefits of CoCounsel Legal and Deep Research is that these features provide embedded guidance to associates as they are drafting documents and conducting research. No matter how much energy library and professional development teams put into training – they can’t provide “on the spot training” for every assignment.  As Agents grow more pervasive and more powerful a key training message will be to remind associates to “check the agent’s work and read every source!”