We require assistants to speak a custom message before triggering a tool. Since we want a variety messages spoken under different cases for our transfer/end call tools, we require the message to be spoken before triggering the tool. I can get the behavior to work as intended in all cases expect for when a customer asked not to be called. __Example where model behaves as intended:__ - When asked to speak to a real person: First, say "Let me get one of our expects on the line.". After, use transfer_call_testing tool Behavior: assistant speaks the message then triggers the transfer_call tool __Example where model deviates from expected behavior:__ - When asked not to be called: First, say "I will add you to our do not call list. Have a great day!". After, use end_call_testing tool Behavior: assistant immediately triggers the end_call tool without speaking the message.

Hi there! I can help ensure your assistants consistently speak the intended message before triggering any tool, including your “do not call” case. I’d start by reviewing your workflow and the conditions around tool triggering to identify why the end_call tool bypasses the message. Could you share how the “do not call” intent is currently configured in your system? @ABC1235

Under our call management settings, we have different criterea for expected outputs based on what the customer says. A few examples from our Call Management Section: ## Call Management - If the conversation goes off-track: "That's an interesting point about [tangent topic], but lets circle back to [relevant qualification topic] first" - If asked to not be called again or to be taken off our calling list: First, say "I will add you to our do not call list. Have a great day!". After, use end_call_testing tool - When asked to speak to a real person: First, say "Let me get one of our experts on the line who can give you the most up to date information". After, use transfer_call_testing tool - For qualified prospects: First, say "Let me get one of our experts online who can give you the most up to date information". After, use transfer_call_testing tool - For unqualified prospects: First, say "It sounds like we don't have any information relevant to you. Have a great day!". After, use end_call_testing tool All of these behave as intended, apart from do not call requests. Any request for do not call will result in immediately triggering the end call tool. Any ideas on why this could be happening? The first thing that came to my mind was there could be some built in logic around do not call requests that disallows the model to speak any further .

Thanks for the detailed examples! I can help you ensure that even “do not call” requests speak the intended message before triggering the end_call tool. The first step would be to review your workflow logic and see if there’s a built-in override for do not call requests. From there, we can adjust the logic or implement a workaround to make message-first behavior consistent across all cases. Can you share how your do not call intent is currently configured? @ABC1235

do you mean for managing opt out requests for for when the model detects that a caller wishes to be opted out?

I can help with both managing opt-out requests and handling detected opt-outs. Let’s discuss privately so I can review your setup and suggest the best fix. @ABC1235

Hi @ABC1235 please watch out for @Vercel . He says he knows the answer to your problem and he can help. But he will ask for money upfront. I made the mistake of doing so and never heard back from him. I already reported him in the # report-scam channel. But I wanted you to be aware of him. Cheers!

thank you for the heads up!