according to gpt i would need at least 10 concurent connections to handle calls per customer this brings the problem that i need to increase the pricing significantly to unprecented heights and unable to use vapi for my voice app