Microsoft used the second day of Build 2026 to introduce Project Solara, an Android based operating system whose central premise is that AI agents, not human authored applications, should be the primary unit of software on a device. The product team is presenting the platform under the more formal name Microsoft Device Ecosystem Platform, and it sits on top of the Android Open Source Project rather than the licensed Google variant, which means the company has full control over the shell, the update path, the identity stack, and the agent runtime layered on top.
The core idea is what Microsoft is calling just in time UI. Instead of designing a screen for each task on each device, Solara expects an agent to assemble an interface at request time, sized and styled for the form factor in front of the user. A meeting summary on a smart display would render as a richer card with controls, while the same summary on a wearable badge might collapse into a single notification with one or two voice actions. Microsoft argues this approach removes one of the most expensive parts of building new device categories, namely the bespoke design and engineering needed to wrap each function in a specialized chrome.
Two concept devices anchored the demo. The first, called Desk, is a conventional looking smart display with a touchscreen, microphones, and a camera, and it can double as a secondary monitor or a thin Windows 365 endpoint. It runs on MediaTek IoT silicon, which is a meaningful signal that Microsoft is willing to build outside the x86 and Qualcomm Snapdragon comfort zone for shared, always on devices. The second is a touchscreen badge worn on a lanyard with 5G, a camera, microphones, and a fingerprint reader. Microsoft positioned it as a way for frontline staff to authenticate to an agent biometrically, record and summarize meetings, and take environmental actions through the camera, all without booting a laptop.
The honest read is that Solara today is closer to a manifesto than a product. Ars Technica, which got an early look at the hardware, reported plainly that none of it works yet, and that Microsoft is committing the spend as part of its broader AI buildout. Partners named so far include AccuWeather, Best Buy, CVS Health, Levi's, and Target, which suggests the company is shaping pilots around retail associates, in store experiences, and health adjacent service desks rather than chasing a general consumer launch.
For technology leaders, three things are worth paying attention to. First, the architectural pattern. Solara is described as chip to cloud, and Microsoft is putting its own enterprise identity, device management, and agent execution containers underneath the AOSP layer. If we already run Intune, Entra ID, and Microsoft 365 Copilot, the integration story for Solara devices will be straightforward, which is exactly how Microsoft historically pulls new form factors into its installed base.
Second, the rebalancing with OpenAI is visible here too. Solara is designed to host multi agent workflows, and Microsoft is making it clear that those agents can be built on its own MAI Thinking models, on OpenClaw, or on third party stacks running inside its execution containers. The platform is being positioned as agent agnostic on purpose, which gives Microsoft a hedge against any single model provider, including OpenAI, deciding to compete more aggressively at the device layer.
Third, the form factor implications for our own roadmaps. Smart badges, shared meeting room appliances, kiosks, and retail surface devices have long been frustrating to build because they each need a custom UI shell. If just in time UI works even partially, the cost curve for a category like a hospital ward badge, a warehouse picker companion, or a branch banking advisor screen could change meaningfully over the next 24 months. We should resist the urge to commit to any single Solara device in 2026, since this is clearly a 2027 and beyond story, but it is worth asking our enterprise architecture group to map which of our existing internal apps could be re expressed as agent capabilities rather than screens.
It is also worth noting Google's parallel direction. At Google I/O earlier this spring, the search team previewed agentic interfaces that build small dashboards and mini apps from a single query, and Android XR continues to lean on generated UI patterns for spatial devices. The convergence is striking. Both hyperscalers are now describing a near future where the operating system's job is to host a fleet of agents and render their output, not to launch icons on a grid. Solara is the most explicit version of that vision from Microsoft so far, and it deserves a place on the watch list of any team planning device or workforce technology investments through 2027.
One final note for buyers. Microsoft has historically struggled to land new device categories, from the Zune through the Surface Duo, and the company's mobile track record is a fair reason for skepticism. The difference this time is that Solara does not require Microsoft to win a consumer device war. It needs to win the enterprise frontline and shared appliance segment where the company already has identity, management, and licensing uses, and that is a far more defensible position than past efforts.
