Carl Pei has never been one to make small claims. The co-founder of OnePlus and current CEO of Nothing — the London-based consumer technology company known for its transparent-backed phones and LED-studded design language — is now staking his company’s future on a prediction that would unsettle every app developer from Cupertino to Shenzhen: AI agents will replace your apps. Not eventually. Soon.
In a recent interview with Digital Trends, Pei laid out his vision for what comes after the app-centric smartphone. “I think in the near future, AI agents will replace a lot of the apps that we use today,” he said. The statement is bold but not entirely original — executives at Google, Apple, Samsung, and a dozen AI startups have all gestured in this direction. What makes Pei’s version interesting is that he’s actually building toward it, and he’s doing so from a position most industry observers would call underdog at best.
Nothing’s approach centers on a project the company has been developing in collaboration with Qualcomm and other partners. The idea is straightforward in concept, staggering in ambition: instead of opening an app to book a flight, order food, or manage your calendar, you’d simply tell an AI agent what you want. The agent handles the rest — interfacing with services, negotiating between platforms, executing tasks across what today requires a half-dozen discrete applications. No more toggling. No more friction between walled gardens.
Pei’s conviction isn’t abstract theorizing. Nothing has already begun integrating AI features into its phones through its Nothing OS, and the company’s sub-brand CMF has been experimenting with more accessible hardware that could serve as delivery vehicles for agent-based interaction. The CEO told Digital Trends that he sees this transition happening within the next few years, not the next decade.
That timeline matters.
It matters because the smartphone industry is stuck. Global shipments have plateaued. The upgrade cycle has lengthened. Consumers don’t see enough reason to buy a new phone every year — or even every two years — because the differences between generations have become incremental. Camera improvements, slightly faster processors, marginally better battery life. The excitement is gone. Pei knows this. He built Nothing partly as an answer to smartphone monotony, using distinctive industrial design and a cult-brand marketing playbook to carve out a niche. But design alone won’t sustain a hardware company forever.
AI agents represent the next potential inflection point. And Pei wants Nothing to be there first — or at least early.
The concept of AI agents replacing traditional apps isn’t new to anyone who’s been following developments at the major platform companies. Google has been pushing its Gemini assistant deeper into Android, with the explicit goal of making it a proactive agent that can take actions on your behalf across apps and services. Apple introduced Apple Intelligence last year with similar ambitions, though its rollout has been slower and more conservative. Samsung’s Galaxy AI features, built partly on Google’s models, have tried to position the phone itself as an intelligent intermediary rather than a dumb container for third-party software.
But here’s where Pei’s argument gets sharper. He’s not just saying AI will make phones smarter. He’s saying the entire app model — the thing that has defined mobile computing since Steve Jobs opened the App Store in 2008 — is approaching obsolescence. Apps as discrete, siloed experiences that users manually open, interact with, and close? That’s a paradigm Pei sees as fundamentally incompatible with how AI works best. Agents need to flow across services. They need to act autonomously. They need to collapse the boundaries between what today are separate applications into a single, continuous interaction layer.
This is a direct threat to the economics that sustain Apple and Google. The App Store and Google Play are toll booths. They extract 15 to 30 percent commissions on in-app purchases and subscriptions. They control discovery, distribution, and monetization. If AI agents bypass apps entirely — if a user never opens the Uber app because their AI agent books rides directly through an API — then the entire revenue architecture of mobile platforms shifts. Dramatically.
Neither Apple nor Google will let that happen without a fight. And this is the central tension in Pei’s vision: can a company as small as Nothing actually lead a transition that the two most powerful platform owners on Earth have every incentive to resist, co-opt, or slow down?
Pei seems aware of the asymmetry. In his comments to Digital Trends, he positioned Nothing not as a platform competitor to Apple and Google but as a company nimble enough to build the right hardware and software integration for an agent-first world. Nothing doesn’t carry the legacy baggage of a massive app store business to protect. It doesn’t have hundreds of billions in services revenue that depends on the current model persisting. That freedom, Pei argues, is an advantage.
Maybe. Or maybe it’s just a different way of describing a lack of market power.
The technical challenges alone are formidable. For AI agents to truly replace apps, they need reliable, real-time access to thousands of services through APIs or other interfaces. They need to handle authentication securely. They need to manage payments. They need to understand context — not just what you’re asking for right now, but what you meant, what you prefer, what you’ve done before. And they need to do all of this without the kind of errors that erode trust instantly. One botched hotel reservation, one wrong medication refill, and users will retreat to the familiar safety of manually tapping through an app.
The AI industry has a term for this problem: reliability at scale. Current large language models, even the best ones from OpenAI, Anthropic, and Google, still hallucinate. They still make confident errors. They still struggle with multi-step tasks that require precise execution across multiple systems. The gap between a demo that looks magical and a product that works reliably every single time is enormous. It’s the gap that has swallowed countless AI startups whole.
Nothing’s partnership with Qualcomm is relevant here. Qualcomm’s latest Snapdragon processors include dedicated neural processing units designed to run AI models on-device, reducing latency and improving privacy by keeping data local. If agents are going to work in real time — booking, canceling, rescheduling, communicating — they can’t afford the lag of round-tripping every request to a cloud server. On-device AI processing is a prerequisite, not a luxury. And Qualcomm, which supplies chips to most of the Android world, is betting heavily that on-device AI will be a key selling point for the next generation of smartphones.
Pei’s timeline of “the near future” is deliberately vague. But the competitive dynamics suggest he can’t afford to wait long. Chinese manufacturers like Xiaomi, Oppo, and Vivo are aggressively integrating AI into their devices and software layers. Samsung continues to iterate on Galaxy AI. Google is rebuilding Android around Gemini. Even startups like Rabbit and Humane tried to build entirely new hardware categories around the agent concept — though both have struggled with execution and market reception, offering cautionary tales about how hard it is to get consumers to abandon the familiar smartphone form factor.
The Rabbit R1 and Humane AI Pin both launched to considerable hype and withering reviews. The hardware was underwhelming. The AI wasn’t reliable enough. The use cases were too narrow. Consumers looked at these devices and asked a simple, devastating question: why wouldn’t I just use my phone? Pei’s answer, implicitly, is that the phone itself should become the agent. Don’t build a new device. Transform the one people already carry.
That’s a smarter bet. But it still requires execution that Nothing has yet to fully demonstrate.
Nothing’s track record is mixed in ways that matter for this discussion. The company has shipped genuinely interesting hardware — the Nothing Phone (1) and Phone (2) earned praise for their design and surprisingly competent cameras at their price points. Nothing OS, the company’s Android skin, is clean and thoughtful. The community around the brand is engaged and vocal. But Nothing remains a niche player. Its global market share is a rounding error compared to Samsung, Apple, or even mid-tier Chinese brands. The company’s last publicly disclosed valuation was around $600 million, which sounds impressive for a startup but is pocket change in an industry where R&D budgets alone run into the tens of billions annually.
So when Pei talks about AI agents replacing apps, he’s painting a picture of a future where Nothing’s small size and design-forward philosophy give it an edge in a world that hasn’t arrived yet. It’s a classic disruptor’s argument: the incumbents are too invested in the old model to move fast enough, and a smaller, more focused company can define the new one.
History offers some support for this view. Apple itself was a niche computer maker before the iPhone redefined mobile computing. Android started as a small startup before Google acquired it. But history also offers plenty of counterexamples — companies that saw the future clearly but couldn’t marshal the resources, partnerships, or distribution to get there before the giants caught up.
Pei’s comments also raise a deeper question about what consumers actually want. The app model persists not just because Apple and Google enforce it but because it works. People understand apps. They know how to find them, download them, use them. The mental model is intuitive. Replacing that with an agent-based system requires users to trust an AI with tasks they currently control directly. That’s a significant behavioral shift, and behavioral shifts in consumer technology take longer than technologists typically predict.
Consider voice assistants. Siri launched in 2011. Alexa in 2014. Google Assistant in 2016. All were supposed to fundamentally change how people interact with technology. And they did — for setting timers, playing music, and checking the weather. For anything more complex, most people still reach for their phone and open an app. The promise of conversational computing has been perennially five years away for over a decade now.
AI agents are more capable than those early voice assistants. The underlying models are orders of magnitude more sophisticated. But the trust gap remains. And trust, in consumer technology, is earned slowly and lost instantly.
Pei deserves credit for thinking beyond the next product cycle. Most smartphone CEOs talk about camera megapixels and screen refresh rates. He’s talking about the structural future of how humans interact with software. Whether Nothing can actually deliver on that vision — with its current resources, its current market position, its current technology partnerships — is the open question.
The answer will depend on execution, timing, and something Pei can’t control: whether the AI models themselves become reliable enough, fast enough, to make the agent-first phone a reality rather than a pitch deck. The technology is advancing rapidly. But rapidly and fast enough are not the same thing.
Nothing’s next major product launches will be telling. If the company ships hardware and software that meaningfully demonstrate agent-based interactions — not just chatbot wrappers on existing apps, but genuine autonomous task completion — it will validate Pei’s thesis in a way that matters to the industry. If it ships another well-designed phone with incremental AI features, the vision stays theoretical.
And theoretical visions, no matter how compelling, don’t ship products.
For now, Pei is doing what he does best: generating conversation, positioning Nothing as the thinking person’s phone brand, and placing a bet on a future that’s plausible but unproven. The smartphone industry needs someone willing to make that bet. Whether Nothing is the company that collects on it is another matter entirely.
WebProNews is an iEntry Publication