About this talk
This talk, presented by Arun Joseph from Exord Systems and Patrick Whelan from Deutsche Telekom, introduces the Agent Definition Language (ADL), a new specification language aimed at enhancing the reliability and scalability of AI agents in enterprise settings. The session discusses the importance of aligning specified behavior with actual agent execution and outlines how ADL achieves this through structured definitions of workflows and validation of tool usage. The language also allows domain experts to define agent skills directly, facilitating easier workflow management without heavy reliance on engineering resources. Additionally, the ADL architecture addresses challenges of scalability by decomposing agent systems into independent units, thereby supporting large-scale enterprise deployments while ensuring model-agnostic execution across various AI technologies.
