​ To learn more, see the sources I considered: [[1] Function Node Overview](https://docs.retellai.com/build/conversation-flow/function-node) [[2] Conversation Flow Overview](https://docs.retellai.com/build/conversation-flow/overview) [[3] Discord message from @jonobeet on 2026-01-11](https://discord.com/channels/1186424376300679209/1294891368505937952/1459795507697291488) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer --- Yes, function nodes can be called with information the AI collected during conversation nodes. Here's how it works: Function nodes are called when entering the node — the function associated with the node will be invoked upon transition into it. [(1) ]() The LLM generates function arguments — the function node settings include an "LLM" option that is "used for function argument generation." This means the AI uses context from the conversation (what it collected in prior nodes) to dynamically populate the arguments for the function call. Transition conditions can route based on collected info — you can write transition conditions on conversation nodes to determine when to move to a function node, and those conditions can evaluate conversation context. [(1) ]() [(2) ]() Response variables can store function results — you can map a function's response to a variable (e.g.,

{{schedules}}

) for use later in the flow. [(3) ]() **Important limitation**: There's a known limitation where full JSON objects from one function node aren't reliably reused as arguments in later function calls, even if mapped and prompted. The workaround is to persist data in flow/global state or flatten it and explicitly bind that variable in the later function's parameter mapping. So in summary: the AI collects info in conversation nodes, then when transitioning to a function node, the LLM generates the function arguments based on that collected context.