At Mobile World Congress this week, AGI Inc. demonstrated the next phase of mobile AI: agents that can move beyond responses and take real actions across apps.
Our focus at MWC was on actions that are truly useful in everyday life. If AI is going to be useful on phones, it must reliably complete real workflows. That requires agents that can navigate app interfaces, coordinate steps, and execute tasks the way a person would. Imagine you give your phone to a friend or executive assistant to do tasks for you while you’re driving. This human-level app usage is our first major goal for mobile action agents.
To demonstrate this, AGI Inc. collaborated with Lenovo on a limited proof-of-concept demonstration within Lenovo Qira. The demo showed how a mobile-use agent can interpret context and execute actions on Android, including booking a restaurant, ordering a ride, and navigating checkout flows, all through the actual UI rather than brittle API integrations.
“AGI Inc.’s technology complemented Lenovo Qira’s Personal Ambient Intelligence vision in this proof-of-concept demonstration at MWC, showcasing how context-aware agents can take coordinated action across apps. Together, we demonstrated how AI can move from passive responses to practical execution.”
Jeff Snow, Executive Director of Product, AI Ecosystem, Lenovo
Reliable mobile-use agents remain the bottleneck for real-world AI adoption on phones. If an agent cannot consistently complete tasks across apps, the experience breaks down. At MWC, we demonstrated a mobile-use agent that works across applications, handles real workflows, and adapts as interfaces change.
While full on-device execution is still ahead, the foundation is in place. As processing moves closer to the device, agents will be able to execute actions with lower latency, can support stronger privacy controls, and fewer dependencies on cloud infrastructure.
Our limited proof-of-concept collaboration with Lenovo illustrates how agent capabilities can integrate with system-level intelligence. Lenovo Qira provide the contextual intelligence and orchestration layer, while AGI Inc.’s mobile-use agent extends that into the action layer on Android devices, showing how AI can move from suggestion to execution across real applications.
MWC marked an important milestone, not just for AGI Inc., but for the broader ecosystem. The path forward for AI on mobile devices depends on agents that can reliably take action, and that is the platform we are building.
This was a controlled proof-of-concept demonstration and not a product announcement.
