Mmarcuscheninvoicelatency.hashnode.dev·5h ago · 7 min readThe caller heard silence for two seconds before the agent spokeA voice agent that felt broken on every first turn, and the latency budget I had to take apart stage by stage to find the two dead seconds. The bug report was one sentence: "callers keep talking ove00
Mmarcuscheninvoicelatency.hashnode.dev·12h ago · 7 min readThe transcript was perfect and the agent still answered the wrong questionThe word error rate was near zero, and the agent kept confidently answering something the caller never asked. The bug was hiding in the punctuation nobody was looking at. The escalated call was a bi00
Mmarcuscheninvoicelatency.hashnode.dev·2d ago · 6 min readThe Friday before we shipped the voice agent, I went looking for 500 callers who did not existThe demo worked. That was the problem. Every time I called our voice agent myself, it behaved. I knew the happy path because I built the happy path. I said my account number clearly, I waited for the 00
Mmarcuscheninvoicelatency.hashnode.dev·Jul 6 · 6 min readThe transfer was warm on our end. The caller didn't feel it.We built a voice agent that escalates to a human queue when it can't close something out: billing disputes, anything with legal language in it, anyone who says "let me speak to a person" twice. The es00
Mmarcuscheninvoicelatency.hashnode.dev·Jun 27 · 3 min readThe 2am call that dropped before the user finished talking, and the week I spent finding out why my tracer never saw itThe call came in at 2am. Not a page, an actual support recording, flagged by a customer who said our voice agent "hung up on her mid-sentence." I pulled the trace. The LLM call was perfect. 380ms, cle00
