avatarin, an AI customer service company spun out of ANA Holdings, built a 24/7 multilingual shopping agent for Yamada Denki using OpenAI's GPT-Realtime. In a two-week public campaign on the retailer's online store, approximately 30,000 people used the agent, and 92% of survey responses came back positive.
"This is not an extension of the conventional chatbot. It is an attempt to expand what customer service can be with AI. We see it as the beginning of an 'interface revolution' in retail," said Akira Fukabori, CEO of avatarin.
From answering questions to understanding people
avatarin had used the OpenAI API for speech recognition, inquiry analysis, and employee training before the Yamada Denki project. Those earlier efforts taught the team what it takes to apply AI in practice. The shopping agent became the first major opportunity to bring those lessons directly to customers.
GPT-Realtime gave the team something they had not found in specialized speech recognition systems: a single model that could work across speech, text, and images with low latency. "Instead of asking people to adapt to the rules of a machine, the technology can adapt much more naturally to the way people communicate," Fukabori said.
A conventional chatbot waits for the right keyword. avatarin's agent listens for context. It responds to uncertainty, shifting preferences, and the details that separate a generic answer from a useful recommendation. Fukabori described the difference with an everyday shopping problem: "Customers do not want a chatbot. They want intelligence. 'I need a refrigerator for a family of four, but my kitchen is small. Which one should I choose?' Real customer service means being able to answer that question."
Giving retail expertise a voice
avatarin designed the agent around three qualities. First, a retrieval-augmented generation system grounds responses in product information while GPT-Realtime keeps the conversation responsive. Shoppers get a fluid voice experience built on relevant product data.
Second, the team translated Yamada Denki's customer service knowledge into the agent's conversation flows and prompting. The system adapts when customers change requirements or move temporarily off topic, while guardrails keep the interaction focused on shopping. Third, the agent takes a proactive approach, asking follow-up questions to uncover actual needs rather than waiting for the next instruction.
OpenAI worked with avatarin to structure complex prompts, optimize API costs for always-on voice service, and share implementation best practices. For teams building similar capabilities, OpenAI Courses cover the tools and techniques behind these kinds of deployments. "I'm excited by how naturally it understands our intentions," Fukabori said, "and by glimpses of intelligence that sometimes go beyond what people can do."
When a conversation becomes an insight
The numbers told one story: roughly 30,000 users, 24/7 multilingual support by voice and text, and 92% positive survey responses. But the deeper discovery was what each conversation revealed. Shopper concerns, hesitations, and decision points surfaced in ways that traditional online shopping had never captured.
"With conventional online shopping, those insights were difficult to see," Fukabori said. "With an AI agent, they become part of the conversation." Customers asked questions after stores closed and spoke candidly about budgets without feeling sales pressure. Each conversation ended with a short voice survey, making feedback a natural extension of the experience. Some customers described the agent as easier to talk to than an actual sales associate.
One Intelligence. One Brand. Every interface.
avatarin's vision extends beyond a single channel. The company wants a single AI agent to serve customers across the web, phone, and physical stores, carrying context forward rather than resetting at each touchpoint. For Yamada Holdings, whose businesses span home appliances, housing, and financial services, that could make AI a consistent companion across daily life.
"AI has clearly shifted into a higher gear," Fukabori said. "I can't wait to see what we can build with this extraordinary intelligence." He added that the company does not want every brand to sound the same: "We want people to experience one intelligence that embodies each company's identity and brand across every interface. The era of single purpose AI is over."
Why this matters for customer support professionals
avatarin's results point to a shift in what customer support teams should expect from AI tools. The agent did not deflect inquiries or reduce interaction to a ticket-it guided 30,000 shoppers through product discovery and purchase decisions with a 92% satisfaction rate. For support leaders, the lesson is that AI for Customer Support can now handle the kind of contextual, follow-up-driven conversations that once required a seasoned associate. The technology exists to build voice agents that ask questions, adapt to uncertainty, and uncover needs rather than wait for a keyword match. The question is whether your current support strategy accounts for that capability.
Your membership also unlocks:
