Hello VAPI Support Team, I’m encountering a problem with the AI Agent when customers provide their phone numbers or emails. If a customer speaks slowly or pauses briefly while giving their contact details, the AI Agent prematurely interprets this as the end of their input and moves on to the next step. This results in incomplete or incorrect information being captured. An example of this issue occurred at reference ID 0e4acfa4-dd63-473b-ba00-5b7c2518eff7. I’d like to improve how the AI Agent handles these situations, possibly by allowing a longer pause before concluding the user has finished speaking. Could you please advise on how best to adjust these settings or implement a fix? Thank you for your assistance.

Hi there! 👋 I understand you're experiencing an issue with your AI Agent prematurely interpreting pauses as the end of input when customers are providing contact details. While I don't have specific information about the reference ID you mentioned, I can definitely help with adjusting your settings to allow for longer pauses. Here are several effective ways to fix this issue: ### 1. Adjust Wait Time Parameters You can modify these key parameters: - **`numWordsToInterruptAssistant`**: Increase this to 3-5 words to give customers more time to speak before the AI can interrupt - **`maxDurationSeconds`**: Set this to 5-10 seconds to define how long the AI waits for a response - **`startSpeakingPlan.waitSeconds`**: Increase from the default of 0.4 seconds to around 0.8-2 seconds to add delay before the assistant starts speaking For example:

javascript
const assistant = {
  // other configurations
  numWordsToInterruptAssistant: 5,
  maxDurationSeconds: 10,
  startSpeakingPlan: {
    waitSeconds: 0.8
  }
};

[Source](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1334027136305991750) ### 2. Enable Smart Endpointing Enable the

smartEndpointingEnabled

feature which uses advanced processing to detect when a customer has truly finished speaking, even with pauses:

javascript
const assistant = {
  // other configurations
  startSpeakingPlan: {
    waitSeconds: 0.8,
    smartEndpointingEnabled: true
  }
};

[Source](https://docs.vapi.ai/customization/speech-configuration) ### 3. Adjust Transcriber Settings For Deepgram, you can set the

endpointing

parameter to a higher value (e.g., 300) to allow for longer pauses:

javascript
const assistant = {
  // other configurations
  transcriber: {
    model: "nova-2",
    language: "en",
    provider: "deepgram",
    endpointing: 300
  }
};

[Source](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1338497725917827196)

Hello brother, this is not something the support can help with, they can only give you advice. The issue is with the prompt and instructions

Hey @GTS, checking if this is resolved for you

Hello brother, this is not something the support can help with, they can only give you advice. The issue is with the prompt and instructions