4 tips for building better AI agents that your business can trust

ZDNET reports on how Thomson Reuters Labs is exploring and deploying AI agents, with CTO Joel Hron outlining four key lessons for building trustworthy agentic systems: measure success with benchmarks and human-in-the-loop evaluation; co-locate experts (designers and data scientists) to improve UX and understanding; decompose and expose proven capabilities as tools for agents; and collaborate externally (Trust in AI Alliance, academic partnerships) to push accuracy and trust. The article highlights products like Westlaw Advantage and the Deep Research agent and emphasizes reaching the final nines of accuracy needed for professional use.