AI agents can now shop, book, and plan on your behalf. So is the future of apps simply to disappear? The reality is more nuanced, more exciting, and more surprising than the headlines suggest.
The Legacy App Shakeout
The vast majority of apps on the App Store today were built in the pre-AI era. Static, single-purpose, often mediocre at serving customer jobs. These legacy apps are the ones that will disappear, and few people will miss them.
But the popular narrative that AI itself will replace the App Store, that ChatGPT becomes the new home screen, is already unravelling. This week, OpenAI killed its Instant Checkout feature, routing purchases back through retailer apps after failing to solve the messy realities of product data, tax compliance, and fulfilment. What died, at least for now, was a specific business model where OpenAI tried to be both the pipe and the toll booth. Consumers rejected the toll booth.
AI isn’t the outcome. It’s a tool to get jobs done. And the organisations that most deeply understand customer jobs and unmet needs are the big consumer brands that have specialised in understanding and serving them for decades. The real shakeout isn’t AI replacing apps. It’s AI-native apps replacing legacy apps. And that story plays out in two simultaneous directions.
The Rise of Domain Apps
On one side, we’ll see a concentration of trusted “Domain Apps”: the very best at serving customers within a sector, leveraging AI and agents to consolidate related jobs-to-be-done into a single, powerful experience.
Imagine if Tesco’s or Walmart’s app became your trusted family nutrition hub, covering meal planning, dietary needs, health goals, and weekly shopping in one seamless experience. Or a hotel app that evolves into a full travel concierge. Or an automotive app that handles everything from browsing to financing to finding a charging station.
These brands already have the deep sector knowledge and customer trust. AI amplifies their advantage. For customers, this means less friction, fewer logins, and genuinely personalised experiences from brands that actually understand what they need.
The brands that don’t achieve domain status? They risk being intermediated and commoditised, or disintermediated entirely by the AI platforms (Amazon, Google, Alibaba) that are building their own commerce intelligence layers. This is a winner-takes-most dynamic within each vertical.
The Simultaneous App Explosion
Here’s the paradox. While consolidation happens at the top, there will simultaneously be an explosion of new apps from below.
The numbers are striking. Since the arrival of agentic coding tools in early 2025, the App Store added roughly 200,000 apps in just three quarters, climbing from 2.08 million to 2.27 million by Q4 2025 (source: 42matters, PocketGamer via Business of Apps). Year-on-year iOS app releases swung from negative growth to nearly 60% by December 2025 (a16z, Sensor Tower, Wells Fargo). The broader vibe coding market is projected to grow at over 30% annually through 2040. Our own expert dev team are leveraging AI to build new apps in hours and have them Apple-approved in a day or two.
This unlocks a long tail of niche, highly specific apps serving previously uneconomical unmet needs: a specialist GLP-1 diet tracker, a local tidal fishing guide, a rare plant identification tool. A few of these may grow to become Domain Apps, but the vast majority will thrive by being brilliant at one very specific job. Think of how YouTube democratised video: most channels are niche, but that’s precisely why the ecosystem works.
What Makes an App AI-Native?
Both trends share a common requirement: intelligence that lives inside the app, on the customer’s own device, not just in the cloud. And there’s a natural progression.
It starts with on-device deterministic intelligence: rules, triggers, and logic. Then probabilistic ML adds contextual personalisation, understanding individual preferences, habits, and intent in real time. And then, naturally, Device Native AI becomes agentic: performing tasks, making recommendations, and orchestrating experiences on behalf of the customer within a trusted domain.
Chat Discovers. Apps Deliver.
AI chat interfaces like ChatGPT and Gemini are powerful for broad, exploratory discovery: “tell me about vacations in SE Asia” or “what’s a good running shoe for flat feet?” But AI-native Domain Apps serve a different purpose. Inside a trusted brand’s app, on-device AI can access far deeper personal context (your health data, your purchase history, your dietary needs, your schedule) to deliver trustworthy, actionable, personalised outcomes. Not generic answers. Your answer.
This is also where trustworthy agentic AI is most likely to emerge for consumers first: not as a standalone bot making purchases on your behalf, but embedded within and extending the customer journeys that brands already provide and customers already trust.
Consider a hotel resort app. Today it handles bookings and room service. As a Domain App powered by on-device AI, it becomes your personal concierge: running with your data, on your device, designed to serve your unique needs at every moment of your stay. It knows your dietary preferences, your schedule, your interests. It suggests, books, and adjusts in real time. Privately. Personally. Proactively. Not a generic chatbot. A trusted agent, operating within a journey you’ve already chosen to be part of.
From One Engine to Many
Just as the single factory steam engine gave way to many smaller electric motors embedded in individual machines, we’re moving from one AI that tries to do everything to a multitude of invisible AIs, each serving specific jobs inside the apps we already trust. As John Dubois at EY recently wrote, the most successful technologies disappear: they weave into the fabric of everyday life until they become infrastructure. That’s exactly where on-device AI is headed.
The Architecture Gap
This is where most organisations stumble. As Bain recently observed, most AI pilots meet expectations but few deliver measurable value, because scaling requires a fundamentally new, integrated architecture. Legacy systems were built for simple request-response functionality. Agentic AI demands systems that support adaptive, multistep, end-to-end actions. The brands that get this right will move AI from a series of experiments to a true operating capability.
That architecture shift is exactly what separates AI-native apps from legacy apps with a chatbot bolted on. And critically, the most trustworthy and capable version of this runs on-device, with personal data never leaving the customer’s phone.
Right Forecast, Wrong Clock
I’ve been thinking about the future of mobile for a long time. Back in 2012, I wrote a series called “The Rise of Mobility and the Disappearing Mobile Phone” which forecast that foldable screens would liberate the smartphone from its rigid form factor, that digital contact lenses would replace handheld screens, and that we were at the neck of the hourglass: functionality concentrated into one device before diffusing back out into a wearable, intra-connected personal area network.
Fourteen years on, much of this has played out. Foldable phones are a commercial reality. Smart watches, AR glasses, intelligent earbuds, and connected rings have all arrived. Hardware functions that collapsed into the smartphone are expanding back out into dedicated wearable devices. But they’re far from ubiquitous, and more advanced concepts like digital contact lenses remain firmly in the lab.
The lesson for anyone forecasting the app landscape today: the direction of travel is usually clearer than the speed. Even when you get the “what” right, the “when” humbles you.
Building for Whichever App Future Arrives First
The future app landscape will simultaneously consolidate and fragment. Legacy apps will fade. Domain Apps will command loyalty. A long tail of brilliant niche apps will serve needs we haven’t imagined yet. And the intelligence infrastructure underneath all of it will progressively move from cloud to device: closer to the customer, more private, more personal.
The direction of travel is clear. The timeline will surprise us. The smartest strategy isn’t to bet on one predicted future. It’s to build adaptable, on-device intelligence infrastructure that serves customers wherever apps, devices, and interfaces go next.
The next few years will be both exciting and unexpected. We’re building the in-app data and intelligence infrastructure to support both sides of this evolution. Happy to connect and share more.