AI agents for business are rapidly becoming a core part of enterprise workflows, but building them without creating trust issues remains a critical challenge. According to Joel Hron, organizations must focus on measurement, collaboration, and proven capabilities to ensure reliable deployment.
As companies adopt agentic systems, AI is shifting from traditional software interfaces to autonomous tools that deliver expertise directly. At Thomson Reuters, AI agents power solutions such as Westlaw Advantage and internal research tools, combining proprietary data with machine learning and large language models to support professional decision-making.
The first priority is establishing clear evaluation frameworks. Hron emphasizes that organizations must define what success looks like through a mix of public benchmarks, internal testing, and human validation. While automated evaluations accelerate development, human experts remain essential to verify accuracy before deployment.
Collaboration between technical and design teams is another key factor. AI agents require a shared interface and understanding between humans and systems. Hron highlights the importance of aligning user experience with agent behavior, ensuring that both operate within a common framework. Close interaction between data scientists and designers helps improve usability and transparency.
Organizations must also focus on extending proven capabilities rather than relying on AI models alone. Instead of expecting agents to handle all tasks independently, companies should integrate existing tools and workflows into agent systems. This approach enhances performance while maintaining reliability across complex use cases.
Finally, building trust requires industry collaboration. Initiatives such as the Trust in AI Alliance, involving organizations like Anthropic, Amazon Web Services, Google Cloud, and OpenAI, aim to improve transparency, explainability, and accuracy in AI systems. Thomson Reuters has also partnered with Imperial College London to advance research in trustworthy AI.
Hron notes that while current models can achieve high levels of accuracy, the real challenge lies in reaching near-perfect reliability. For enterprises, achieving this level of precision is essential to building trust and ensuring adoption.
As AI agents become more embedded in business operations, organizations that prioritize evaluation, collaboration, and transparency will be better positioned to deploy trustworthy systems at scale.
重要なポイント:
- AI agents for business require strong evaluation and validation frameworks.
- Human oversight remains critical for ensuring accuracy and trust.
- Collaboration between technical and design teams improves usability.
- Integrating proven tools enhances agent performance and reliability.
- Industry partnerships support transparency and trustworthy AI development.
ソース：
https://www.zdnet.com/article/4-tips-for-building-better-ai-agents-business-can-count-on/
