Traditional function calling approaches for LLM-powered conversational agents suffer from unpredictable execution, poor debugging capabilities, and the "prompt and pray" problem that kills development velocity. This talk introduces Process Calling - a superior paradigm where LLMs invoke stateful, multi-step processes rather than atomic tools.

Unlike function calling's stateless operations, Process Calling enables agents to maintain context and execute deterministic business logic across conversation flows. Using real production examples, we'll show how Rasa's CALM framework implements Process Calling to build conversational agents that actually work in customer-facing scenarios. Come learn why the future of customer facing agent development isn't about better prompts—it's about better abstractions.