AI Retail Agent 24/7: The Yamada Denki Case with GPT-Realtime

The timeline: from zero to 30,000 interactions in fourteen days

The project arises from the collaboration between avatar, a Japanese startup specializing in telepresence and physical agents, and OpenAI. The goal was precise: to eliminate downtime in the customer service of Yamada Denki stores. Yamada Denki is one of Japan's largest consumer electronics chains, with a diverse clientele in terms of language and age group.

Avatar has integrated GPT-Realtime — OpenAI's model optimized for low-latency voice conversations — within a physical agent deployed in retail locations. The result is documented directly by OpenAI in its official case library: 30,000 interactions in the first two weeks, with a satisfaction rate of 92% among those who responded to the post-interaction survey.

Additionally, the system handles requests in multiple languages without manual configurations for each. This feature is central: Yamada Denki serves an increasing number of tourists and foreign residents. A multilingual agent reduces reliance on specialized staff and lowers scaling costs.

Solution Architecture: What Makes GPT-Realtime Suitable for Physical Retail

GPT-Realtime is designed to handle streaming audio with lower latency than standard models. In a retail context, every second of perceived waiting time impacts the experience. Therefore, choosing a real-time model is not aesthetic; it's functional for conversational fluidity.

The Avatarin agent combines three technological layers. The first is the language model, which interprets the request and generates the response. The second is the physical presence layer—the robot or terminal—which gives the agent a visual identity. The third is the system that integrates with product data and shelf availability information.

Indeed, an in-store conversational agent without access to real inventory would be useless. The real technical complexity lies not in the language model, but in the integration with existing management systems. This is the hurdle that any retailer must face before considering such a deployment.

To further explore AI architectures applicable to the Italian context, the team SHM Studio — AI Services Provides a preliminary assessment of necessary integrations.

Winners and losers: who really profits from this model

The most obvious winner is the retailer. A 24/7 agent eliminates the cost of nighttime or holiday staff for basic information functions. Furthermore, it scales without significant marginal costs: the fiftieth simultaneous user costs the same as the first.

However, there are less obvious losers. The first is the retailer that adopts the technology without a clear customer journey strategy. A poorly configured agent—one that provides generic responses or is disconnected from actual inventory—creates frustration rather than satisfaction. Yamada Denki’s 92% is not a matter of chance: it is the result of careful design.

The second potential loser is unqualified sales staff. The AI agent handles standard information requests. Consequently, the value of human staff shifts towards complex consulting and closing sales. Those who do not evolve their roles risk progressive marginalization.

Finally, there is a third scenario: the competitor that doesn't move. In mid-complexity retail—electronics, furniture, do-it-yourself—the one that first implements a reliable AI agent builds a perceptual advantage that is difficult to recover.

SHM Studio's reading: real scalability or isolated case?

The question Italian marketing managers ask themselves is legitimate: can a Japanese case study be replicated in Italy? The answer is complex.

On one hand, the underlying technology is identical. GPT-Realtime is available via OpenAI's API for any developer or agency. Therefore, the technical barrier is not geographical. On the other hand, the context is different: Italian clientele has specific relational expectations, and an agent perceived as cold or mechanical can damage the brand instead of supporting it.

We of SHM Studio We observe that the successful cases share a common element: experience design precedes technology selection. First, it's defined what the agent should do, what questions it should be able to handle, and what tone it should use. Only then is the model selected and the integration built.

According to research from McKinsey on the economic potential of generative AI, retail is one of the sectors with the highest value that can be captured by conversational automation. However, capturing that value requires integration with existing processes, not just installing a chatbot.

Operational implications for Italian retail: three concrete scenarios

For marketing managers evaluating a similar initiative, it is useful to distinguish three levels of implementation maturity.

• Basic Level — Digital Information Agent: A chatbot on an e-commerce website or app, powered by GPT-Realtime or equivalent models, that answers questions about availability, shipping, and return policies. Low investment, measurable impact in a few weeks. Suitable for retailers with existing structured digital traffic.
• Intermediate level — omnichannel agent: The same agent integrated between digital channel and physical point of sale, with real-time inventory access. Requires integration with ERP or POS. The Yamada Denki case falls here.
• Advanced Level — Proactive Agent: the system not only responds to requests but anticipates needs based on browsing behavior or purchase history. This level requires a quality database and a strategy of digital marketing integrated.

For each level, the critical factor is not the AI model but the quality of the product data and the consistency of the communication tone with the brand. An agent who responds in a way that is inconsistent with the retailer’s positioning creates cognitive dissonance in the customer.

Metrics to monitor after go-live

Yamada Denki's 92% satisfaction metric is an output KPI. But process KPIs are just as important for understanding whether the agent is truly performing well.

First, the containment rate: the number of requests resolved by the agent without human escalation. A rate below 60% indicates problems with training or scope definition. Second, the Average resolution timemust be lower than that of the equivalent human channel, otherwise the perceived value is null.

In addition, it is essential to monitor conversational abandonment rate: how many users abandon the session before receiving a response. A high rate indicates perceived latency or irrelevant responses. Finally, the post-interaction conversion rateIn retail contexts, the agent must contribute to sales, not just provide information.

To structure an effective measurement system, the Google Ads campaigns and the activities of SEO they can be integrated with agent data to build a consistent funnel from acquisition to conversion.

What no one tells you: the problem with cultural context

The Yamada Denki case takes place in Japan. This is not a neutral detail. Japanese culture has a very high threshold for technological acceptance, especially for robots and automated agents in public spaces. The average Japanese customer is accustomed to interacting with automated systems in contexts where an Italian customer would expect a human being.

Therefore, replicating the case in Italy requires a change management phase with the customer, not just with internal staff. The launch communication must be carefully managed. The agent must be presented as an additional service, not a replacement. The tone must be warm, not technocratic.

According to Harvard Business Review, the negative perception of AI agents in physical contexts is often linked not to the technology itself, but to the lack of transparency about the automated nature of the system. Openly stating that it is an AI agent increases trust, not decreases it.

This is a principle that we at SHM Studio we also apply it to digital content and communication design: transparency is a competitive asset, not a weakness. For those who want to delve deeper into how to build an effective narrative around these initiatives, the service SEO copywriting and that of LinkedIn campaign they offer concrete tools for communicating innovation to stakeholders and B2B customers.

Next moves: what to consider in the next six months

The Avatari n-Yamada Denki case is a useful benchmark, not a model to be copied. Each retailer has a different context: customer mix, catalog complexity, internal digital maturity.

So, the first recommended step is an audit of the current customer experience: where are repetitive requests concentrated? Which questions receive slow or inconsistent answers? Which channels are unattended during off-peak staff hours? These three questions identify the correct scope for an initial AI agent.

Subsequently, the prototyping phase should be brief and measurable: two to four weeks, one channel, a limited set of intents. Yamada Denki's time-to-value—two weeks—is not a technological miracle. It is the result of a well-defined scope from the outset.

For those who want to explore these opportunities with the support of a specialized team, the starting point is a consultation with SHM Studio. We support marketing teams from strategy definition to go-live, with a focus on measurability at every stage. Further resources and use cases are available in the SHM Studio Blog and in the section web development for those evaluating deeper integrations with existing systems.