In late 2022, the launch of ChatGPT marked a significant turning point in the landscape of artificial intelligence, particularly for companies like Thomson Reuters. This event was not merely a technological milestone; it served as a catalyst that validated the aspirations of engineers and product leaders who had long envisioned leveraging AI to tackle complex customer challenges that previous generations of technology struggled to address. The emergence of agentic AI represents a paradigm shift from traditional reactive systems, such as chatbots, to proactive agents capable of interpreting complex objectives, planning multistep tasks, and executing workflows alongside human professionals.
Traditional chat-based interfaces have proven useful for reactive tasks, but they often fall short in aligning with user goals, managing multistep reasoning, or taking meaningful action. These systems can resemble teammates waiting for instructions rather than anticipating needs and proactively addressing them. In contrast, agentic AI systems are designed to act rather than merely respond, marking a new chapter in the evolution of AI in professional domains.
The development of agentic AI has necessitated platforms that prioritize trust, accuracy, and domain expertise—values that Thomson Reuters has diligently integrated into its systems over the years. The company has introduced agentic capabilities in products such as CoCounsel Tax, a specialized AI for tax and accounting, and CoCounsel Legal, which is recognized as the legal industry’s first professional-grade agentic AI research tool. These innovations signify a commitment to enhancing the capabilities of professionals in high-stakes fields where precision and reliability are paramount.
CoCounsel is engineered to interpret complex objectives, plan and execute multistep workflows, and deliver results with the same level of precision as seasoned lawyers, accountants, or compliance officers. It leverages industry-leading tools, including Westlaw for legal research, Practical Law for procedural guidance, and Checkpoint for tax and accounting expertise. This enables CoCounsel to perform real work, such as calculating figures to complete a tax return, rather than simply providing information.
The implications of this technology are profound. Professionals can now delegate entire assignments—such as drafting and reviewing legal motions or working through a tax return—confident that the work will meet industry standards. This capability is underpinned by extensive expert guidance from thousands of legal, tax, and compliance specialists, ensuring that outputs are not only technically accurate but also aligned with the practices of experienced practitioners.
As David Wong, Chief Product Officer, and Joel Hron, Chief Technology Officer at Thomson Reuters, observe, this transformation is particularly significant in high-stakes domains like law, tax, compliance, and risk management. In these industries, accuracy, transparency, and trust are non-negotiable. AI must perform reliably, align with regulatory requirements, and support nuanced human judgment. The shift from merely providing answers to delivering outcomes is a critical evolution in the role of AI in professional settings.
Three years ago, Thomson Reuters’ AI journey looked markedly different. The company was an early adopter of tools like GitHub’s Copilot, and today, more than 80% of its engineers utilize such tools weekly. However, this was just the beginning. The focus has now shifted toward developing full agentic systems—tools capable of referencing internal documentation, interacting with servers, retrieving live data, triaging and fixing bugs, and even building applications from scratch.
Unlike software code, which typically has testable and verifiable outcomes, many expert domains present a range of acceptable answers, some of which are better than others. This variability underscores the importance of human judgment in the process. To facilitate this, Thomson Reuters has embedded AI engineers within domain expert teams, fostering collaboration that accelerates iteration and builds trust. With over 250 AI engineers working alongside more than 4,500 domain experts—including lawyers, accountants, and compliance leaders—the company is shaping AI into practical, real-world capabilities.
Imagine a system designed for lawyers that goes beyond merely suggesting clauses in contracts. Such a system could compare documents, identify legal risks, and escalate complex issues for expert judgment. Similarly, a tool for tax professionals could flag compliance risks, adapt to real-time data, and complete multistep workflows, moving beyond the simple retrieval of tax codes. These concepts are not abstract; they represent embedded, outcome-focused systems that are beginning to redefine how professionals operate.
However, the advancement of agentic AI brings with it a set of challenges. The tech culture has historically celebrated the mantra of “move fast and break things.” Yet, in fields like law and tax, breaking things is not an option. While speed is important, trust is even more valuable. No matter how advanced agentic capabilities become, they will not be adopted if professionals cannot trust them. Intelligent systems must transcend mere results to provide transparency, consistency, and oversight.
Designing effective agentic products is as much a human challenge as it is a technical one. Agentic systems must understand when to escalate decisions, how to explain their reasoning, and how to adapt without deviating from the user’s standards. Human-in-the-loop controls are essential to this process. Experts guide development, stress-test edge cases, and ensure performance in the contexts that matter most.
When systems are empowered to plan and act, small misalignments in goals, context, or data quality can lead to significant errors. Overreliance on automation without clear guardrails can sideline human judgment when nuance is critical. Some of the most powerful features of agentic systems are invisible, such as their ability to maintain context, reason over multiple sources, or decide when not to act due to insufficient information. These capabilities enable professionals to work with greater confidence.
In their leadership roles, Wong and Hron emphasize the importance of adaptability. They value curiosity, the ability to learn quickly, and cross-disciplinary collaboration over specialization alone. The organizational structure has been reconfigured to support small, highly aligned teams, empowering subject matter experts to shape AI behavior. This approach allows for rapid iteration without sacrificing rigor or trust.
The future of true agentic capabilities is not about creating the fastest or most autonomous system. Instead, it is about building the most useful one—a system that can reliably assist professionals in high-stakes moments where time is of the essence. The most meaningful agentic AI will expand what professionals can achieve, particularly when margins for error are slim.
As the landscape of AI continues to evolve, the emphasis on trust, accuracy, and domain alignment will remain paramount. The integration of AI into professional workflows is not merely about enhancing efficiency; it is about fundamentally transforming how professionals engage with their work. By prioritizing these values, Thomson Reuters and similar organizations are paving the way for a future where AI serves as a trusted partner in navigating the complexities of law, tax, compliance, and beyond.
In conclusion, the rise of agentic AI signifies a profound shift in the relationship between technology and professional practice. As organizations embrace these advancements, the focus must remain on building systems that not only demonstrate intelligence but also foster trust and reliability. The journey toward fully realizing the potential of agentic AI is ongoing, but the commitment to enhancing professional capabilities while maintaining the highest standards of accuracy and integrity will undoubtedly shape the future of work in significant ways.