How avatarin built a 24/7 retail agent with GPT-Realtime

Source: OpenAI Date: 2026-07-30 URL: https://openai.com/index/avatarin

Summary

An OpenAI customer case study: Japanese company avatarin built a 24/7 retail agent on the GPT-Realtime API. Source page 403’d on direct fetch and secondary coverage doesn’t add operational detail beyond the headline — no disclosed metrics on call volume, resolution rate, or deployment scale.

Implications

Enterprise proof-point for voice/realtime agents in a physical-retail setting, feeding the same OpenAI-as-platform motion as the academic-researcher program this week — model launches increasingly paired with named customer deployments rather than benchmarks alone. Thin signal on its own (no hard numbers); mainly useful as evidence GPT-Realtime is landing continuous, unattended production use rather than staying a demo-only API.

(Source page returned 403 on direct fetch; summary grounded via secondary coverage and search results, not the primary page.)