Deepblue Technologies launched Wowoker, a large-model AI-agent work phone, at a product event on October 29, Shanghai Securities News and Jiemian News reported. Built with Huawei's HarmonyOS ecosystem, it targets government enterprises, state-owned firms, finance and the broader enterprise market. Several tech outlets described it as the world's first large-model agent work phone — a claim from the company, not yet independently verified.
The hardware base comes from Huawei: Kirin chips, device hardware and HarmonyOS form the terminal layer, fused with Deepblue's multimodal large model, its Decision Model and an AgentOS orchestration layer. According to the company, the product is positioned to let agents execute business tasks directly — packing lead generation, customer outreach, sales conversion and customer management into one device under a "hardware + SaaS + data + AI" architecture, with on-device models assisting in social communication scenarios.
Putting large models in phones is not new; most flagships ship on-device models. What is new is the product logic: rather than a general assistant for consumers, this terminal acts as an execution unit of enterprise systems. Outbound calls, marketing distribution, multi-platform operations and customer-data capture — actions usually scattered across separate SaaS tools — are wrapped into agent tasks that flow automatically on the phone, concentrating management and compliance boundaries at the terminal.
For Huawei, it extends the HarmonyOS ecosystem further into enterprise AI endpoints; for Deepblue, moving from algorithms to integrated software-plus-hardware is a route to deliverable, billable products. The open question is execution quality: business tasks demand reliability, privacy and auditability far beyond chatbots, and every step of an agent completing a sales cycle needs to be logged and controllable.
The enterprise-agent competition is shifting from chat boxes into business systems. Wowoker's answer is to claim the terminal entry point first — whether it works will show in the task-completion rates of the first government and enterprise customers.