Home / Technology / AI Agents: Your New Work Collaborator?
AI Agents: Your New Work Collaborator?
21 Mar
Summary
- AI agents will soon impact all professional roles.
- Thomson Reuters uses in-house and off-the-shelf tools.
- Building trustworthy AI requires evaluations and human input.
AI agents are set to fundamentally alter professional landscapes, with organizations soon adopting these technologies through various means. Thomson Reuters Labs, under CTO Joel Hron, is at the forefront of integrating generative AI, machine learning, and agentic systems. The company utilizes a strategic mix of in-house developed models and readily available commercial tools, augmented by its own extensive knowledge base.
Hron emphasizes four crucial lessons for developing reliable agentic AI. Foremost is the critical need for robust evaluations to define and measure success. This involves utilizing public benchmarks for initial insights, creating custom internal benchmarks with detailed evaluation criteria, and crucially, maintaining human expert review before product deployment.
Furthermore, a deep understanding of agent operations and seamless integration with the user experience are paramount. This requires fostering a common language and interface between human users and AI agents, enabling transparency into their respective thought processes. Cross-functional collaboration between business teams, designers, and data scientists is essential for this synergy.
Despite advancements in AI capabilities like code writing, planning, and reasoning, agents are not omniscient. Their potential is significantly amplified when integrated with existing software tools, allowing them to leverage proven functionalities. Thomson Reuters is actively adapting its decades-old applications into tools that agents can effectively utilize, thus extending model capabilities.
To champion trustworthy AI, Thomson Reuters co-founded the Trust in AI Alliance, a forum for industry leaders to share insights on engineering trust into agentic systems. This collaborative effort, alongside academic partnerships like the one with Imperial College London, focuses on achieving the incremental accuracy gains—the "last two nines"—that are critical for market competitiveness and professional confidence in legal, tax, and compliance applications.
