Japanese Retailer Deploys GPT-Realtime Agent to 30,000 Shoppers in Two Weeks
The short answer
Yamada Denki's AI customer-service agent handled 30,000 interactions in its first two weeks, with 92% of surveyed users reporting positive experiences — one of the first publicly documented retail deployments of OpenAI's real-time voice API at meaningful scale. Retail customer service automation is now being measured in weeks to deploy, not quarters.
- retail
- customer service
- voice AI
- GPT-Realtime
- multilingual
- automation
What happened
Avatarin, a robotics and remote-presence company, deployed an AI agent at Yamada Denki — one of Japan's largest electronics retailers — using OpenAI's GPT-Realtime API. The agent provides 24/7 multilingual customer support to shoppers in-store. Within two weeks of launch, 30,000 people had interacted with it, and 92% of survey respondents reported positive experiences.
Multilingual support was a core requirement: Yamada Denki serves both domestic and international customers, and the agent handles conversations across languages without additional staffing. The deployment is notable because the numbers come from a live retail environment, not a controlled pilot.
What it means for your business
Retail customer service automation is now being measured in weeks to deploy, not quarters. A 92% satisfaction rate from real shoppers gives you a benchmark that's more useful than vendor marketing claims. If your business handles repetitive customer inquiries — store hours, product availability, how-to questions — this case is worth studying before you budget for additional support staff. The 24/7 multilingual angle is particularly relevant if you serve customers in more than one language and can't afford round-the-clock staffing. Costs and setup complexity for GPT-Realtime integrations vary widely, so get a concrete quote before assuming this is out of reach.
Source: OpenAI