Based on the booking issue you described, there are several configuration areas that commonly cause booking problems on Retell AI.
Timezone and Current Time Configuration
The most common cause of incorrect booking behavior is timezone misalignment. You should include the {{current_time}} dynamic variable in your prompt before triggering availability checks and booking functions.(1) This ensures the LLM knows the current date and time when making booking decisions.(1)
Make sure the timezone in your Retell function call matches the timezone configured in your calendar system (e.g., Cal.com).(1) Timezone mismatches can cause the agent to check availability for the wrong day or offer slots that don’t actually exist.(1)
Dynamic Variables for Booking Functions
If you’re experiencing issues where the agent isn’t using correct information during booking, ensure you’re including the retell_llm_dynamic_variables object in your request payload.(2) This was identified as a critical fix for dynamic variables working correctly.(2)
Prompt Configuration
Your prompt should explicitly ask for the user’s timezone before checking availability and booking.(1) Additionally, structure your availability prompts to be clear about time slots. For example, mention only the start time of slots and specify that each slot has a defined duration.(1)
Debugging Booking Issues
You can review what the LLM sent during failed bookings by:
- Going into the call transcript
- Looking at tool calls where you’ll see the tool result highlighted in a different color
- Clicking on the tool result to see what was sent
- Clicking on the tool name to see what the LLM sent(1)
Common errors include “Time range is in the past” which indicates the LLM doesn’t properly understand the current time context.(1)
Platform Reliability
Retell AI typically achieves sub-600ms to 620ms average end-to-end latency with optimized configurations.(3) The platform uses a proprietary turn-taking model with predictive timing to ensure smooth transitions.(3) Issues like the one you’re experiencing are typically configuration-related rather than platform reliability issues.
I recommend checking your timezone settings, adding the {{current_time}} variable to your prompts, and reviewing your tool call logs to identify exactly what the LLM is sending to your booking function.
Would you like help reviewing your specific prompt configuration or booking function setup?