Voice AI isn't a technical problem. It's a psychology problem disguised as a technical problem. Everyone in this space is chasing models. Which LLM are you using. What's your speech-to-text provider. How fast is your inference. And look, that stuff matters. But it's maybe 20% of what makes a voice AI experience work. The other 80%? User experience. Not UX the way most people think about it. After 18 months building voice AI systems, I started seeing it. Voice AI UX is kind of a seven-layer problem. Most people only think about one or two. Layer 1. Latency. If the response takes more than 800ms, people hang up. Not a preference. A reflex. Layer 2. Voice realism. If the tone doesn't fit the industry... a dental office getting the same voice as a debt collector... people feel it right away. They might not know why. They know something's off. Layer 3. Conversational psychology. Where most voice AI falls apart. Humans don't list six time slots and ask you to pick one. They say "mornings or afternoons?" and narrow from there. Psychology thinking, not engineering thinking. Layer 4. Error recovery. Bad voice AI says "sorry, I didn't catch that." Good voice AI says "I want to make sure I got that right... did you say Thursday?" One sounds like a robot. The other sounds like someone paying attention. Layer 5. Interruption handling. Real people talk over each other all the time. If your voice AI can't handle someone cutting in mid-sentence and respond to what they said... the whole thing breaks. Layer 6. Builder experience. If it takes a developer and three weeks to set up, you've lost most of the market. Plug and play for non-technical people. That's its own UX problem. Layer 7. Dashboard. Once the agent is live, you need calls, bookings, conversions, drop-offs... all in one place. If the data lives in five tools, you don't have a dashboard. You have a scavenger hunt. Seven layers. And I haven't touched handoff UX, post-call follow-up, or context persistence. The cool thing is... when all seven line up, the experience feels invisible. The caller doesn't think about the AI. They think "that was easy." I'm turning this into a full playbook. "The 7 Layers of Voice AI UX." Each layer broken down with principles, examples, and what to test for. Drop "UX" in the comments and I'll send it over.