### Answer ## Using Global Nodes for Data Corrections For allowing users to change previously captured information at any point in the flow, the recommended approach is using a Conversation Node → Logic Node pattern with global nodes [(1) ]() . The best approach is to use a Global Conversation Node connected to an Extract Variable Node (or Logic Node) . This pattern works because it allows the agent to listen for correction requests at any point in the conversation, then extract the updated data . To implement this: - Set up a Global Conversation Node that listens for correction keywords or phrases - Connect it to an Extract Variable Node to capture the updated information - Use the "return to previous node" functionality to resume the flow after the correction ## Alternative Approaches Without Global Nodes ### 1. Flex Mode Enable Flex Mode at the component or agent level [(2) ]() . When enabled, all nodes get converted to a single flex node, allowing the agent to switch context between different tasks dynamically . This means: - Users can complete multiple tasks at the same time - The agent can move to the proper task if the user jumps ahead - After switching context, the agent can resume the previous task without repeating completed steps ### 2. Make Node Instructions More Flexible According to the debug guide, you can make node instructions more flexible and let the model handle the details naturally [(3) ]() . This allows the conversation to adapt when users deviate from the defined flow . ### 3. Add More Transition Conditions Review your transition conditions and make them more flexible to handle edge cases where users want to go back or

skip steps .

​ To learn more, see the sources I considered: [[1] Discord message from @mauvatar on 2026-02-19](https://discord.com/channels/1186424376300679209/1186462578310910112/1474156743444660314) [[2] Flex Mode](https://docs.retellai.com/build/conversation-flow/flex-mode) [[3] Debug guide](https://docs.retellai.com/build/conversation-flow/debug-guide) ​

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

@AI Evy how can I tell the difference between a user restating what they said previously versus actually making a correction?

### Answer I don't have specific documentation about distinguishing between user restatements versus corrections in the available sources. The sources cover conversation flow nodes, code execution, turn-taking mechanisms, and custom LLM integration, but don't explicitly address detecting the semantic difference between a user repeating information versus correcting it. This distinction would typically be handled through: - Careful prompt engineering in your conversation nodes to recognize correction keywords ("actually," "I meant," "no, it's," etc.) - Using dynamic variables to track what was previously captured and comparing new input - Implementing logic in code nodes to detect mismatches However, these specific implementation patterns aren't detailed in the available documentation. If you'd like, I could try to provide a general approach based on conversational AI best practices, but it wouldn't be directly from Retell AI documentation. Would that be helpful?

message has been deleted

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