Warm or cold transfer, just takes too long to hand over. The team is telling me it rings once by them and many times the agent just hangs up on the caller…
I’ve changed the model to 4.1. I’ve extended the wait for human and ring time so at least the calls won’t disconnect, but the callers are agitated and many just put down.
Call ID: call_a6319e220b659491d207a2496cf my test call
Call ID: call_a3da815f4520d9c9ca9c3f6d875 really bad. And no sound though it was set to ringtone.
Call ID: call_a3da815f4520d9c9ca9c3f6d875 in this one, the team answers and I’m not sure what happened.
Team checked the the Call IDs and Recommended fixes (in order of impact):
Disable human detection for this transfer. In your transfer node settings, set “Is there any internal queue or hold system before an agent answers?” to no. Since you’re always transferring to a staffed dispatch center (not a personal cell where voicemail might answer), you can safely skip human detection entirely. This will eliminate the detection delay and the misclassification hangups — the AI will immediately start the handoff message as soon as the call connects.
Your IVR is the biggest time contributor (~15s). If possible, configure your phone system to bypass the IVR for calls from your Retell number (+13237364441) — this alone would cut the transfer time roughly in half.
First, thanks. I’ turned off wait for human but that’s really causing a mess.
When I transfer back to Ring Central it goes into a queue and by the time the staff picks up the Agent’s already provided the warm transfer details to no one. call_8969be2b36b25c3695205f5b020 - same message twice, what’s happening here?!
Team checked this and looks like the original detection failures were caused by RingCentral’s IVR announcements confusing our detection model, and it seems they need the setting back on for their use case.
Re-enable “Wait for human” on your transfer node (set “Is there any internal queue or hold system before an agent answers?” back to its previous setting)
Keep the timeout at 51 seconds — this gives enough time for your dispatcher queue to connect someone
Remove the duplicate confirmation: Your flow currently tells the caller “connecting you to dispatch” in two places (the global prompt AND the collect_route_stop node). Remove one so the caller doesn’t hear it twice.