How avatarin built a 24/7 retail agent with GPT-Realtime
- ID
- 9311
- Status
- summarized
- Published
- 30 Jul 2026, 8:00 AM
- Fetched
- 31 Jul 2026, 9:45 AM
- Provider
- OpenAI News
- Category
- ai-labs
- Original URL
- https://openai.com/index/avatarin
- Source URL
- https://openai.com/news/rss.xml
Summary
- Score
- 3.5
- Created
- 31 Jul 2026, 4:35 PM
- Tags
- Audience
- ai_agent_userssaas_founders
What happened
OpenAI publishes a customer story where avatarin, an AI customer service company spun out of ANA Holdings, built a 24/7 multilingual retail shopping agent called Kurashi-Marugoto for Yamada Holdings using OpenAI's GPT-Realtime API. The agent engaged 30,000 shoppers during a two-week public experience with 92% positive survey responses.
Why it matters
This is a vendor case study, not independent reporting. The only actionable detail for builders is that GPT-Realtime was used to power a voice-based, multilingual retail agent at scale—useful as a reference architecture if you are evaluating realtime voice APIs for customer service, but no pricing, latency, or failure-mode data is provided to make a real build-vs-buy decision.
Discussion angle
What's missing from this case study that you'd actually need before choosing GPT-Realtime for a production voice agent—cost per conversation, latency, fallback handling, and how '92% positive' was measured.