Based on the documentation, I can identify the core issue and provide a solution.
The Core Problem
Your Begin node is configured with “User speaks first”, which causes the agent to wait for the IVR to finish speaking before transitioning.(1) Press Digit nodes evaluate whether to press a digit each time the user finishes speaking, and this timing is also affected by the detection delay setting.(1)
The critical issue is that transitions in Press Digit nodes occur only when the user is done speaking.(1)(2) This means your Begin node won’t transition until the IVR stops speaking, which may be too late for time-sensitive IVR menus.
Recommended Solution: Use Logic Split Node Immediately
The Logic Split node transitions immediately when the agent enters it.(3) The time spent in this node is minimal, and the agent does not speak.(3)
Restructure your flow like this:
Begin Node (User speaks first)
|
v
Logic Split Node (immediate evaluation)
|
+-- If {{ivr_variant}} = "custom_press_digit"
| |
| v
| Press Digit Node
| (with Detection Delay configured)
|
+-- Else
|
v
Other handling nodes
Key Configuration Steps
1. Configure Detection Delay in Press Digit Node
Some IVR systems speak slowly. Set a delay on pauses to ensure the whole IVR menu is captured before the agent makes a decision.(1) The documentation recommends setting this to 1 second.(1)
2. Write Clear IVR Navigation Instructions
Provide clear instructions so the agent knows whether and what digit to press.(1) Include keywords or phrases to listen for, as well as which ones to avoid.(1)
Example prompt structure:
Your goal is to reach the scheduling or appointments department.
Preferred navigation keywords:
• Scheduling
• Appointments
• New patients
• Front desk
Avoid:
• Billing
• Referrals
• Medical records
• Clinical departments
If you are unsure which IVR option is correct:
Choose the option most closely related to scheduling or appointments.
3. Configure Transition Conditions
Write conditions like “Reached scheduling department” or “Menu repeated 3 times” to handle transitions.(1) Ensure you cover both successful navigation and potential failure scenarios or edge cases.(1)
Why This Works
The Logic Split node evaluates conditions immediately upon entry and branches to the corresponding destination nodes without the agent speaking.(3) This eliminates the delay caused by waiting for the IVR to finish speaking in the Begin node.
Once in the Press Digit node, the detection delay setting gives the IVR time to finish speaking before the agent evaluates what digit to press.(1)
Additional Considerations
If you’re still experiencing issues after implementing these changes, consider:
- Adjusting the LLM temperature if responses are inconsistent(4)
- Adding transition finetune examples to help the model understand your expectations(4)
- Making node instructions more flexible to handle edge cases(4)
**