Are we actually ready for agentic AI smartphones?

• ByteDance’s agentic AI smartphone sold out in hours, triggers backlash.
• Consumers like the idea of AI autonomy more than the reality.
• High-autonomy AI weakens consumer trust.

When ByteDance’s agentic AI smartphone prototype sold out in hours of launch on December 2, it seemed like validation of consumer appetite for autonomous mobile assistants. Then came the backlash.

Videos demonstrating the ZTE Nubia M153’s ability to autonomously book restaurants, make purchases, and edit photos sparked immediate privacy concerns – forcing ByteDance to reportedly scale back capabilities before wider release.

The contradiction reveals a fundamental tension in AI development: technological capability has outpaced human readiness. As Asia-Pacific leads the world in AI adoption, the question isn’t whether the technology works, but whether we’re psychologically and practically prepared for smartphones that act without constant permission.

Asia-Pacific’s AI enthusiasm meets autonomy anxiety

The region’s enthusiasm for AI is undeniable. According to Boston Consulting Group’s July 2025 survey of over 4,500 employees in nine Asia-Pacific markets, adoption rates reach 92% in India, with optimism about AI highest in China at 70%, Malaysia at 68%, and Indonesia at 69%.

Asia-Pacific is projected as the fastest-growing market for AI agents globally, with the region expected to grow at the highest compound annual growth rate of 19.8% through 2034. Yet this adoption appetite doesn’t automatically translate to comfort with autonomous systems.

Research published in Information Technology & People found that AI autonomy moderates the relationship between consumer trust and AI service adoption. Specifically, high-autonomy AI weakens the positive link between trust in a company and willingness to adopt AI services – a finding that was replicated in 23 companies and six industries in a representative survey.

The ByteDance prototype exemplifies high autonomy: an AI agent with operating-system-level access that can execute multi-step tasks, access apps, process payments, and manipulate data without waiting for explicit approval at each step.

This represents a qualitative leap from voice assistants like Siri or Google Assistant, which operate in defined boundaries and require frequent confirmation.

What research reveals about human-AI trust

Multiple studies confirm that perceived autonomy – not just technical capability – shapes adoption. A 2024 study in AI & Society examining trust in generative AI found that fairness significantly enhances trust, but transparency and accountability alone don’t guarantee it.

Social presence and emotional intelligence positively impact trust, suggesting humans respond better to AI that feels collaborative rather than controlling.

The KPMG 2024 Generative AI Consumer Trust Survey of 1,000 US consumers found that 51% are “extremely or very” excited about generative AI, and 42% believe it significantly impacts their personal lives now. However, when AI systems demonstrate high autonomy – the ability to make decisions and take actions independently – trust requirements fundamentally shift.

A study in Consumer Research on consumer autonomy in AI-assisted decision-making found that when people feel they retain ultimate authority in decision-making, they’re more likely to embrace AI services. The key concern isn’t whether AI can perform tasks, but whether users maintain control.

The ByteDance phone’s immediate privacy backlash stemmed precisely from this: consumers saw an AI with privileges that appeared to override their control.

The daily use reality check

The practical question facing agentic AI smartphones is whether daily users will accept autonomous agents handling routine tasks. Deloitte’s survey of over 11,900 individuals in 13 Asia-Pacific economies found that generative AI saves daily users an average of 4.4 hours at work and 4.5 hours at university. The time-saving benefits are measurable and significant.

However, 64% of employees believe their tasks will be automated or augmented in the next five years, while simultaneously expressing concerns about whether businesses are taking full advantage of AI responsibly.

The top three barriers to business adoption include lack of talent, concerns about risk, and insufficient understanding of the technology – barriers that parallel consumer concerns about autonomous smartphones.

IDC research indicates that approximately 70% of Asia-Pacific organisations expect agentic AI to disrupt business models in the next 18 months. Yet the gap between organisational adoption and consumer comfort remains significant.

McKinsey’s survey found that 23% of organisations are scaling agentic AI systems somewhere in their enterprises, with an additional 39% experimenting – but most are doing so in controlled business functions, not consumer-facing applications with direct personal impact.

Where agentic AI smartphones stand today

The technology exists. ByteDance’s Doubao large language model already serves 159 million monthly active users as a chatbot and productivity tool. Operating-system-level integration enabling autonomous task execution is technically feasible, as the M153 prototype demonstrated.

What’s missing is the governance framework that enables user trust. Research shows that successful AI adoption requires transparency about what the AI can access, clear boundaries on autonomous actions, comprehensive logging of all activities, and straightforward mechanisms for users to override or revoke permissions.

Anthropic’s enterprise AI solution demonstrates these requirements: role-based access, centralised provisioning control, audit logs for compliance monitoring, and training on internal knowledge. Consumer applications of agentic AI smartphones will need equivalent frameworks – privacy-by-design, not privacy-as-afterthought.

The regional context matters. Asia-Pacific’s “hands-off” regulatory approach to AI, as documented in the Southeast Asia Public Policy Institute’s analysis, prioritises boosting investment and capabilities over immediate regulation. This creates space for innovation but also means consumer protection mechanisms lag behind technological deployment.

The path forward

The ByteDance-ZTE experiment offers valuable lessons. Consumer curiosity exists – the immediate sell-out proves demand. The subsequent privacy panic reveals that curiosity alone doesn’t equal readiness. The gap between these two responses defines the current state of agentic AI smartphones: technologically possible, psychologically premature.

For agentic AI smartphones to move beyond prototype status to daily use, several conditions must align. Users need a transparent understanding of what autonomous agents can access and do. Clear, intuitive controls must enable users to define boundaries and revoke permissions easily.

Comprehensive audit trails should document all autonomous actions, creating accountability without constant interruption. Asia-Pacific’s leadership in AI adoption positions the region to define how this technology matures.

With India’s 92% AI adoption rate and China’s 70% optimism about AI, the enthusiasm exists. The challenge is channelling that enthusiasm toward autonomous systems that feel collaborative rather than controlling – AI that augments human decision-making rather than replacing it.

The ByteDance prototype’s trajectory from sell-out to scaled-back suggests the market is delivering feedback. The question isn’t whether agentic AI smartphones will arrive, but whether they’ll arrive with the trust mechanisms that make daily use comfortable rather than concerning. Until then, the gap between capability and comfort remains the defining constraint on autonomous mobile AI.

Want to learn more about Cloud Computing from industry leaders? Check out Cyber Security & Cloud Expo taking place in Amsterdam, California, and London. This comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.