KDDI and AVITA have formed a partnership to develop and deploy humanoid robots in commercial customer service settings, with trials at physical retail locations scheduled to begin in Autumn 2026.
The collaboration targets a specific operational gap: frontline service roles that require nonverbal communication, empathy, and physical presence. Standard industrial automation handles repetitive tasks well, but breaks down when confronted with unexpected equipment failures, complex customer interactions, or situations that demand social cues like eye contact and facial expressions.
How the Humanoid Works
The hardware draws on a concept model designed by Hiroshi Ishiguro. The humanoid features a compact skeletal structure approximating a typical Japanese physique, with silicone skin and specialized mechanical systems that produce warm facial expressions synchronized directly with spoken dialogue. Embedded camera sensors track moving objects to generate natural eye contact. Quiet pneumatic actuation produces fluid movement with natural “micro-variations,” avoiding the rigid, mechanical quality that has historically made robots feel out of place in hospitality environments.
That design specificity matters. Customer-facing deployment fails or succeeds on whether people feel at ease, and the engineering here is oriented entirely around that problem.
The Network and Compute Infrastructure
Operating a humanoid in a live commercial space demands continuous, low-latency data transmission. KDDI supplies that communications backbone, enabling real-time remote control and cloud-based processing of the visual and motion data generated during customer interactions. That data then feeds back into the AI system to improve behavioral precision and expand autonomy over time.
To handle the computational load, the two companies plan to use GPUs hosted at the Osaka Sakai Data Center, which began operations in January 2026. They are also exploring integration with an on-premises deployment of Google’s Gemini generative AI model, keeping sensitive data processing secure while supporting complex dialogue requirements.
Building on Earlier Work
This is not a cold start. KDDI and AVITA previously collaborated on a remote customer service platform using digital avatars, deployed at retail locations including Lawson and au Style shops. Moving from screen-based avatars to free-moving physical units is a direct extension of that earlier infrastructure and commercial experience.
Deployment at au Style shops will be among the options considered once trials begin. The progression follows a clear logic: enterprises that already run digital avatar programs have the network foundations, data governance workflows, and customer interaction data needed to onboard physical humanoids more effectively.
What This Signals for Enterprise Adoption
Shrinking labor pools are accelerating interest in physical AI, particularly in countries like Japan with acute demographic workforce pressures. But the ROI case depends on identifying where nonverbal, empathetic engagement is genuinely necessary rather than simply automating whatever is easiest to automate.
Governance frameworks will need to evolve alongside deployment. Visual and motion data collected in physical spaces introduces customer data management obligations that go beyond what digital-only AI programs typically require.
Trials in Autumn 2026 will test whether the hardware performs at the standard the design ambitions suggest.
Photo by Kindel Media on Pexels
This article is a curated summary based on third-party sources. Source: Read the original article
