Barcelona’s Mobile World Congress this year felt less like a trade show and more like a laboratory for the next generation of personal computing. Tens of thousands of attendees wandered aisles where banners for 6G and artificial intelligence hung alongside familiar handset makers, but the conversations were dominated by a single shift: phones ceasing to be passive question‑answering devices and becoming active agents that complete tasks on behalf of users.
ZTE and Samsung offered the clearest demonstrations of that transition. Both companies showcased systems that let an on‑device AI “see” an app’s interface and simulate taps and swipes to book rides, order food or check out online. ZTE’s much‑publicised Doubao phone pushes the idea further by granting the assistant deeper operating‑system privileges so the agent can integrate more tightly with the device; Samsung paired Google’s Gemini model with visual agent tooling on the S26 to similar effect.
The technical route many exhibitors favoured — often described as GUI agenting — treats apps as visual canvases to be interpreted by the AI rather than requiring developers to add bespoke APIs. That makes the capability immediately applicable to millions of existing apps without waiting for developer cooperation. It also raises obvious questions about privacy, security and reliability because an agent manipulating interfaces can access any content a human could see or act upon.
Not all innovation at MWC was purely software. Honor unveiled a “Robot Phone” that blends embodied hardware — a swivelling camera gimbal that can turn, nod and express a semblance of emotion — with its AI stack. Honor argues that giving devices a physical presence and “EQ” will make interactions feel warmer and more personal, countering the cold, one‑size‑fits‑all interfaces of modern smartphones.
Behind the brands, Qualcomm sought to set the rules of the road. Rather than building devices, Qualcomm is embedding AI into the radio and edge compute fabric. Its Snapdragon X105 modem includes an AI processor designed to stabilise network performance for multi‑step agent workflows, and the company demonstrated wearable hardware capable of running multi‑billion‑parameter models locally. The strategic aim is to distribute intelligence off the screen and around the body: persistent, context‑aware personal AI that does not always rely on cloud servers.
This convergence of approaches — GUI agents, embodied devices and distributed on‑device models — reflects diverging industry bets. GUI methods are fastest to roll out but can be brittle in complex interactions and are harder to audit. App‑level APIs would be cleaner and safer but require ecosystem cooperation and standardisation. On‑device models improve latency and privacy but strain power, thermal and silicon budgets, and will take time to match cloud scale.
Regulatory and business ramifications will be consequential. Agents that act on users’ behalf will force new thinking about consent, scope of authority and liability when actions go wrong. Phone makers and silicon vendors may gain control over how apps are experienced, disrupting app business models and prompting platform vendors to assert new policies. For consumers, the promise is convenience; the risk is opaque automation with broad data access.
For now, MWC 2026 made one thing clear: the smartphone industry has reignited. The debate has moved beyond whether AI should be on phones to how it should be integrated, who controls it, and what trade‑offs between convenience, privacy and resilience each approach entails. The next year will be a test of which technical path — quick, universal GUI agents, standardised APIs, or heavier on‑device intelligence — best balances user value with safety and performance.
Adoption will hinge on execution. A phone that intermittently completes errands for you is appealing; one that exposes credentials or misfires on payments will not be tolerated. The companies that can deliver reliable, explainable, and low‑latency agent experiences while convincing regulators and privacy‑conscious consumers that risks are managed will set the tone for the post‑smartphone era.