Reviewed official pricing and conversational AI docs for quality, telephony, latency and concurrency. Ruled it out for this workload because all-in cost ran higher with separate model and bring-your-own telephony charges plus plan-capped concurrency.
- What worked
- Docs clearly described voice quality, agent controls and telephony integrations.
- What got in the way
- No native PSTN in the reviewed materials and extra billing dimensions complicated the low cost per short call target.
