# Warm transfer human detection not working, causing downstream conversation bugs **URL:** \ **Category:** Support Help\ **Tags:** bug\ **Created:** [June 13, 2026, 3:29am UTC](https://community.retellai.com/t/warm-transfer-human-detection-not-working-causing-downstream-conversation-bugs/3091 "2026-06-13T03:29:39Z")\ **Posts on this page:** 4\ **Page:** 1
**Author:** ![zachsents](https://yyz2.discourse-cdn.com/flex008/user_avatar/community.retellai.com/zachsents/32/1859_2.png) [@zachsents](https://community.retellai.com/u/zachsents)\ **Post date:** [June 13, 2026, 3:29am UTC](https://community.retellai.com/t/warm-transfer-human-detection-not-working-causing-downstream-conversation-bugs/3091/1 "2026-06-13T03:29:39Z")
Hi, We’re using Conversation Flow transfer nodes with transfer\_option.type: “warm\_transfer” and human detection enabled. In this case, the transfer destination connected to voicemail, but Retell did not appear to classify it as voicemail/no-human quickly. Instead, the caller sat in the transfer flow until the full human-detection timeout elapsed. Reference call: call\_cfb29c050d5f170d267e888bda0 What happened from the transcript/timeline: The agent initiated a warm transfer to +15035725477 with: - transferRingDurationMs: 35000 - agentDetectionTimeoutMs: 40000 - optOutHumanDetection: false From the logs: 15:31:57.409 — transfer\_call tool called 15:32:29.885 — transfer call connected 15:32:29.885 — “Start detecting human speech” 15:33:10.370 — transfer result returned: “Call connected but no human response detected.” 15:33:10.436 — flow transitioned to our Unavailable node So after the transfer leg connected, Retell spent about 40.5 seconds in human detection before failing, which matches the configured agentDetectionTimeoutMs. It looks like the failure happened by timeout, not because Retell recognized voicemail early. The destination audio was clearly voicemail: “You’ve reached Brent McGraw… I can’t get to your call right now… Please leave me a message…” followed by voicemail instructions. Afterwards the transfer, the agent also fired off 2 responses back-to-back. Thanks. ---
**Author:** ![mark1](https://yyz2.discourse-cdn.com/flex008/user_avatar/community.retellai.com/mark1/32/2524_2.png) [@mark1](https://community.retellai.com/u/mark1)\ **Post date:** [June 13, 2026, 1:01pm UTC](https://community.retellai.com/t/warm-transfer-human-detection-not-working-causing-downstream-conversation-bugs/3091/2 "2026-06-13T13:01:58Z")
Hello @zachsents I have escalated your call id to the team for a review. Thank You ---
**Author:** ![mark1](https://yyz2.discourse-cdn.com/flex008/user_avatar/community.retellai.com/mark1/32/2524_2.png) [@mark1](https://community.retellai.com/u/mark1)\ **Post date:** [June 15, 2026, 1:15pm UTC](https://community.retellai.com/t/warm-transfer-human-detection-not-working-causing-downstream-conversation-bugs/3091/3 "2026-06-15T13:15:39Z")
Hello @zachsents You’re correct in your analysis. On the warm-transfer detection leg, voicemail and IVR classifications are produced but don’t fast-fail the detection process. Only a confirmed human response resolves it early, which is why the leg ran the full 40 seconds before returning “no human detected.” This is a platform-side gap, not a config issue on your end. Filed with engineering. In the meantime, lowering `agentDetectionTimeoutMs` will reduce the wait, though it increases the chance of a false-negative on slow-to-answer humans. We will update you once engineering has a fix in place. Thank You ---
**Author:** ![zachsents](https://yyz2.discourse-cdn.com/flex008/user_avatar/community.retellai.com/zachsents/32/1859_2.png) [@zachsents](https://community.retellai.com/u/zachsents)\ **Post date:** [June 16, 2026, 7:07pm UTC](https://community.retellai.com/t/warm-transfer-human-detection-not-working-causing-downstream-conversation-bugs/3091/4 "2026-06-16T19:07:01Z")
Thanks for the reply, I’ll lower the timeout a bit for now. Can you keep me in the loop about the engineering changes?