Jun 12, 2025 | agentic AI

The last decade of rapid AI innovation has progressed faster than the language we use to describe it. From predictive analytics and machine learning to early-stage AI bots like IBM’s Watson, and now the rise of generative and agentic AI, the landscape has evolved so quickly that clearly defining it can be challenging. Along the way, many have been eager to capitalize on each new buzzword as an opportunity to make their technology sound cutting-edge, even if it does not truly meet that standard. 

This AI-washing phenomenon is most pronounced now in the world of agentic AI, where many companies are competing to develop AI-powered solutions that not only generate content such as text, images, or code based on prompts but also interact dynamically within a given workflow, making decisions and adapting based on feedback. At the most basic level, these agentic models enable more complex problem-solving by planning, reasoning, and reacting to data inputs. However, that base level definition is really just a starting point. When true agentic AI models are fully integrated into complex workflows, they enable the performance of complex multi-step assignments, transforming AI from a mere tool into an intelligent partner. 

Reinventing Workflows with Agentic AI 

Let’s look at some examples that help illustrate the difference between the two. Today, all the research capabilities we offer in solutions like AI-Assisted Research in Westlaw or Checkpoint Edge are using agentic AI models to refine search results, weed out unnecessary information, and surface highly curated, accurate, and explainable results faster than ever before possible. These solutions are not only responding to natural language prompts; they are assessing the relative value of different results, making decisions about relevant precedents and dynamically cross-referencing with new information as it becomes available. While that’s a great step forward from the old days of labor-intensive information gathering using Boolean search terms, it is still more of an enhancement to an existing workflow than a wholesale reinvention of the process. 

The real breakthrough innovations Thomson Reuters is now enabling with agentic AI go several steps further than that by embedding AI agents at multiple points in a complex workflow to not only refine results, but partner with users to help them do the work. Our recently launched CoCounsel Tax, for example, is capable of preparing and reviewing returns, automating diagnostics, even engaging with client documents and escalating when professional judgment is required.   

Creating a Native AI Experience 

This is what we refer to as a “native AI experience.” It is not just enhancing the old way of doing things. It is fundamentally reframing the way the task is completed by taking a tax return from a scanned document to a structured review to recommended tax strategies for clients in one seamless, completely automated flow, without requiring users to proactively move in-and-out of multiple different applications or dig around for information. And, because the system is grounded in the Checkpoint database and the IRS code, its results are iron-clad accurate. 

Similarly, we’re developing agentic AI capabilities that deliver what we’re calling “Touchless Compliance,” meaning our global trade management solutions are able to automate each step from determination to e-invoicing, right through to compliance filing. And, coming soon to our Legal solutions, our agentic AI models will be able to analyze facts, summarize documents – even pull in relevant precedent from Westlaw or Practical Law as needed. Importantly, these solutions do not just retrieve information; they reason, adapt and move the task forward to execute complex, multi-step processes. 

Informing the Way Forward 

It’s innovations like these that make the definition of true agentic AI so much harder to conceptualize and to achieve for businesses that do not have the robust proprietary datasets and deep experience reinventing professional workflows that we have cultivated over the last 150 years at Thomson Reuters. It’s relatively easy to apply the latest off-the-shelf model to an ad-hoc task like research or auditing and call it agentic AI. But the real role for agentic is so much bigger. The solutions we’re developing today with agentic AI are completely reinventing the way enabling functions like tax, trade, legal, HR, procurement, risk and compliance professionals do their jobs, giving them a leg up on their competition and helping them produce a quantity and quality of work that was not possible a few years ago. 

For more information about Thomson Reuters corporate solutions, please visit https://www.thomsonreuters.com/en/products-services/corporate-solutions