ByteDance’s unveiling of an agentic AI smartphone prototype with ZTE on December 2 drew fast excitement from consumers, followed almost immediately by questions about privacy. The rush of attention pushed the device to a quick sell-out, then forced ByteDance to scale back several features. Beneath the noise, though, the launch points to a larger shift: AI smartphones that can run system-level agents may soon play a central role in how companies manage work, automate tasks, and support employees in the field.
The ZTE Nubia M153, powered by ByteDance’s Doubao model, is more than a flashy gadget. It offers an early look at how agentic AI could reshape productivity if companies can trust devices that act on their own and handle multi-step tasks without constant supervision.
From curiosity to potential workplace tool
Consumers were drawn to features like hands-free restaurant reservations, automatic photo adjustments, and quick price comparisons across apps. For businesses, the appeal is different. Gartner expects that by 2028, a third of enterprise software applications will include agentic AI, up from less than one percent in 2024. Since smartphones remain the main device in many workplaces, they serve as a practical testing ground.
In fields like manufacturing, construction, healthcare, and energy, AI agents could support safer decisions and take over routine procedures. Nicholas Muy, CISO at Scrut Automation, says the benefits for frontline teams are clear but warns that early adopters must watch for errors and security gaps. Research from McKinsey shows that about a quarter of companies are already using agentic AI in at least one area, with nearly four in ten experimenting. Unlike consumer use, enterprise deployments require guardrails such as logging, compliance checks, and role-based permissions. Those were missing in ByteDance’s consumer-first prototype.
Isaac Sacolick
China’s integration advantage
ByteDance chose to partner with ZTE instead of designing its own hardware. The strategy mirrors successful approaches from companies that treat AI as a system-level feature rather than a device feature. Doubao already holds a strong lead in China’s consumer AI market, with 157 million monthly users as of August 2025. Tencent’s Yuanbao trails at 73 million.
This approach also gives ByteDance a path into second-tier manufacturers that want AI capabilities without building their own assistants. Large phone makers, including Apple and Huawei, already have the resources to develop proprietary agents, which limits ByteDance’s prospects in the top tier. For companies managing phones across global teams, this raises both options and challenges: more flexibility in hardware, but also more complexity in managing data, policies, and compliance.
Privacy fears revealed enterprise expectations
Much of the backlash came after entrepreneur Taylor Ogan shared viral demos showing the device performing tasks with deep system access, including payments and data transfers. What looked impressive to some immediately raised red flags for others. The concern was simple: who controls the agent?
In a survey by Forum Ventures of 100 enterprise IT leaders, trust topped the list of adoption barriers. Jonah Midanik, a general partner at the firm, noted that agentic AI works on probabilities, not certainty, which creates hesitancy in regulated environments. ByteDance’s decision to scale back features showed a recognition that enterprise-grade AI must include fine-grained permissions, activity logs, and clear limits on what the agent can do.
Forum Ventures.avif)
Enterprise use cases differ from consumer needs
For frontline workers, agentic AI could surface equipment histories during repairs, suggest safer routes based on real-time conditions, or walk users through complex procedures without searching across multiple apps. Healthcare teams could access patient data and decision support more quickly. Finance professionals could receive automated, compliance-checked guidance.
According to PwC, almost eight in ten organisations have already adopted AI agents in some form, and nearly all IT leaders plan to expand their use in 2025. Yet a survey from Cloudera shows that success depends on industry-specific data pipelines, clear decision processes, and slow, controlled rollouts.
The consumer phone market is moving fast too. IDC expects 912 million generative AI-enabled smartphones to ship by 2028. Still, business users focus on traceability and security rather than convenience.
Global competition shapes adoption
Apple’s delayed release of Apple Intelligence in mainland China has left a temporary opening for ByteDance and other Chinese AI leaders. Apple’s tight integration of hardware and on-device processing still appeals strongly to enterprise buyers who prioritise privacy. ByteDance, meanwhile, aims to scale quickly by licensing Doubao across many manufacturers. This could create early market standards in China that are difficult for Western rivals to match.
For global companies, the result is a fragmented device landscape. Counterpoint Research reports that Asia-Pacific is the fastest-growing region for AI agents, while the United States currently accounts for just over 40 percent of revenue. Enterprises may have to manage separate device strategies to meet different regional rules and data policies.
What enterprise leaders should look for next
ByteDance’s prototype highlights three requirements companies should expect when considering agentic AI smartphones:
• governance frameworks that define what an agent can or cannot do and record every autonomous action
• hybrid AI approaches that balance on-device processing with cloud-based reasoning to meet regional compliance needs
• phased deployments that start with low-risk workflows and expand as systems prove reliable
Solutions like Anthropic’s enterprise platform show what this might look like in practice, with features such as central provisioning and full audit trails. Amazon’s use of AI agents to modernise Java applications is one example of how companies can find value while limiting exposure.
The ByteDance–ZTE phone may be a consumer product on the surface, but it previews a future where agentic AI is baked into every smartphone. As Gartner predicts that 15 percent of work decisions will be made autonomously by 2028, companies face a choice: shape this transition early or adapt later as consumer devices push boundaries.
Enterprises that prioritise security, transparency, and clear governance from the start are the ones most likely to benefit as agentic AI moves from hype to everyday utility.
