Agentic AI Use Cases

Agentic AI systems rely on large language models to plan and execute multi-step tasks by interacting with tools and collaborating with other agents, creating new demands on Internet protocols for interoperability, scalability, and safe operation across administrative domains. This document inventories representative Agentic AI use cases and captures the protocol-relevant requirements they imply, with the goal of helping the IETF determine appropriate standardization scope and perform gap analysis against emerging proposals. The use cases are written to expose concrete needs such as long-lived and multi-modal interactions, delegation and coordination patterns, and security/privacy hooks that have protocol implications. Through use case analysis, the document also aims to help readers understand how agent-to-agent and agent-to-tool protocols (e.g., [A2A] and [MCP]), and potential IETF-standardized evolutions thereof, could be layered over existing IETF protocol substrates and how the resulting work could be mapped to appropriate IETF working groups.