After demoing 18 platforms in 90 days, here are the 12 questions that actually separate the production-ready vendors from the demos.
I demoed 18 voice agent platforms in the first 90 days of 2026. Twelve of them looked identical in the demo. Five broke the minute I gave them real call volume and a real integration. One shipped a prompt change to production on a Friday night and took a client offline for six hours. Here is the buyer's guide I wish I had had.
The stack we actually use Below is the production stack we ship for this category. This one ships in 72 hours and survives a real-world Monday morning. Pipecat for voice, a thin Express service for chat, deployed on Fly.io or Render. We avoid Vercel for anything stateful. Default to Claude 3.7 Haiku for the chat layer and Claude 3.7 Sonnet for the planning and tool-use steps.
After demoing 18 platforms in 90 days, here are the 12 questions that actually separate the production-ready vendors from the demos. It is filed under AI Voice because that is where operators looking for this problem actually start, and it is written from production work rather than from a content calendar.