Lenovo's announcement of Qira at CES 2026 is not just another assistant launch. With Qira, Lenovo expresses where personal computing is headed: from application hopping and rapid repetition toward a layer of ambient intelligence that persists across devices, transfers context, and takes action.
Lenovo describes Qira as “personal ambient intelligence” that appears as Lenovo Qira on PCs and Motorola Qira on phones, embedded at the system level and designed to be “always present” with user consent (see Lenovo press release: Introducing Lenovo and Motorola Qira).
Who Qira is trying to become
The most interesting part of Qira is that it is presented as intelligence that runs across devices, rather than as a single app. Lenovo bases its impressions on three attributes assigned to specific moments:
- Presence
- Activities
- Perception
Think about behaviors like Next move (proactive suggestions), Write for me (inline editorial office), Live interaction (multimodal screen/camera cooperation), Catch me (post-departure summaries) i Watch out for (transcription, translation, summaries).
These may be known use cases for AI, but have not been presented in this way before. Lenovo understands that AI must evolve from “talking about anything” to “doing something real.” To facilitate this evolution, Qira's design language is explicitly agentic: it orchestrates work between applications and devices, coordinates specialized capabilities, and reduces the amount of human-mediated coupling required to maintain context.
Why Qira's idea is more important than its list of features
Qira looks like a product, but behaves like a hypothesis: the operating system becomes less important than the intelligence layer running above it. This idea appears in my analysis of the agent operating system, the move from files and applications to intent and orchestration as the primary interface.
For context, my deeper look at how agent layers can turn Windows, macOS, and Linux into compatibility shells is here: Agent operating system at SeriousInsights.net and Why Windows Just Got Ruined in the Age of Agent OS on a tech perspective.
Qira also points out something I've been following since the announcement of the AI computer: An AI computer cannot be defined solely by a neural processing unit (NPU). The real question is what does a primarily AI experience do when you can combine voice, image, documents and workflows – and when you can combine it on a laptop and phone without falling apart.
This question has been front and center at Serious Insights since early PC AI conversations: Thinking out loud: What should an AI computer do?.
Instead of throwing their hands in the air as Dell did recently, on PC Gamer (Dell's CES 2026 chat was the nicest AI briefing I've had in maybe 5 years), on an AI computer, as the hardware becomes available to the consumer, Lenovo has doubled down on efforts to implement essential features that even Microsoft leaves on the table.
Hybrid AI, but with a partner-like ecosystem
Lenovo has no doubt that Qira is a hybrid. On-device processing is set as default for privacy and latency reasons, and cloud services extend capabilities as needed. This architecture matters because it is where economics, management and trust collide.
Lenovo has also telegraphed where it wants the ecosystem to run, listing partners: Microsoft for local-cloud coordination, Stability AI for generating on-device images in the Creator Zone, Notion for workspace inference, Perplexity for source research, and Expedia/Vrbo for travel-intent flows.
While this is a practical set: productivity substrate, creation, knowledge store, research and services, it is not exhaustive. For Lenovo to help buyers make a meaningful shift to AI, it will need to ensure its AI offerings are integrated with all popular apps, services and tools.
The partner list also reveals an underlying design tension. The personal ambient agent needs a unified memory to feel coherent. Unified storage requires access to the same data that enterprises spend a fortune trying to manage. When these worlds come together, identity, consent, retention policies and auditability become product features rather than compliance afterthoughts.
The hidden cost is still work
Qira is positioned as continuity: fewer thread interruptions due to re-entry of prompts and device switching. This is exactly the kind of promise that helps mitigate the tax on AI, alleviate the extra effort that goes into preparing, verifying, correcting, and integrating results.
An overview of the AI tax can be found here: AI tax.
Ambient intelligence reduces friction when it works. It also makes bugs seem more invasive because the system is present everywhere. Once the agent is empowered to take action, the failure mode changes from “wrong answer” to “wrong move.” Therefore, agent security is no longer just a topic of scientific articles; this is a product requirement.
A practical list of agent failures in real-world deployments can be found here: How agents can make a mistake.
From an enterprise perspective, readiness is not measured by the number of co-pilots deployed. Readiness is measured by whether an organization can operationalize consent, identity, knowledge hygiene, and continuous measurement without turning every workflow into a compliance effort.
When implementing Qira, Lenovo must remember all this.
Ambient AI still runs on watts and politics
It's easy to talk about personal AI as if it exists in a frictionless cloud. This is not the case. The hybrid architecture described by Lenovo is a response to two constraints: the need for privacy and responsiveness, and the reality where computing is expensive, energy-intensive and increasingly geopolitical.
In an interview with Foxit, I discuss governance, energy and geopolitics: “The future of artificial intelligence is not what you think” (Foxit interview).
The practical implication is that Qira's success depends on the mess in between: what can be done on-premises, what requires cloud acceleration, and what needs to reach third-party models, and how these decisions translate into corporate policy and regional regulations. Product teams must treat “where the intelligence runs” as a design parameter, not just an implementation detail.
What to watch as Qira grows
Lenovo says Qira will roll out to select Lenovo devices in the first quarter of 2026, and will later expand to supported Motorola smartphones, with over-the-air updates for existing Lenovo AI Now users. Deployment strategy matters because agents working across devices only seem real when they are always present, and AI consistency is difficult to scale.
Three things will determine whether Qira becomes a defining layer or a smart moment at CES:
- Trust mechanics: clear and transparent control of what Qira remembers, where it stores it and how it uses it – in relation to the identity of consumers and enterprises.
- Action boundaries: visible barriers and reversibility of agent actions, especially when Qira coordinates third-party services.
- Knowledge quality: less hallucinations and greater provenance when Qira draws from personal data, workspaces and internet research.
If Lenovo gets this right, Qira will become less about competing with the chatbot and more about changing expectations about what the device ecosystem can do and how we will work in the future.
