We are connecting AI to a conference, and it's crucial that the AI connects at a specific moment with minimal lag. Currently, we call the VAPI SIP phone, and it takes time until the connection is established, the AI is initialized, and it begins speaking. Is there a way to reduce this latency? For instance, it would be ideal to establish the connection, pre-initialize the AI, and have it remain silent until we make an API call to prompt the AI to start speaking.