Tech Leaders Discuss Defining Boundaries for Autonomous AI Agents

Key summary

Executives from Omise, Microsoft, and OpenAI shared strategies for managing AI agent autonomy, risk assessment, and accountability at a forum hosted by Agoda in One Bangkok. The discussion highlighted that while AI agents can handle coding, testing, and customer interactions, organizations must clearly define which tasks allow autonomous decision-making versus human oversight. Key challenges include ensuring humans understand system actions to approve them effectively, distinguishing between low-risk reversible tasks suitable for automation and high-impact irreversible tasks requiring human review, and maintaining the ability to audit agent actions and data sources. The panelists emphasized that delegating work does not absolve organizations of responsibility; clear internal roles must be assigned for development and testing teams regardless of AI involvement.