Where the workflow shifted
OpenAI Realtime API and the voice agents guide place low-latency voice, tool use, and session state into the same product-design surface.
Teams should define what callers can actually finish: pricing questions, booking, rescheduling, order lookup, lead capture, human handoff, and confirmation.
Tool names are not outcomes
The signal matters when it changes how a team ships, reviews, or recovers work, not when it only names another tool.
Check permissions and failure
- List the top three call intents and map each to knowledge, required fields, confirmation points, and failure paths
- Keep the test narrow: one low-risk task or tool entry before connecting permissions, logs, failure handling, and human takeover to production
What still needs proof
Optimizing only for natural speech leaves real callers stuck on rescheduling, pricing, and handoff tasks. Keep the original source open so the announcement, the evidence, and this site's interpretation stay separate.