There was a time when every capability we needed existed as a separate, physical thing. If you wanted to listen to music, you carried a Walkman. If you wanted to take a photo, you picked up a camera. Navigation meant unfolding a paper map and, if you were careful, carrying a compass. Each object had a single, well-defined purpose. The world was simple, but rigid. Every new need meant acquiring and carrying another device.
Then came the smartphone, and with it, a profound shift. Capabilities detached from physical objects and became software. One device could now do everything. The camera became an app. The map became an app. Music, banking, communication, travel: everything became an app. It felt like liberation. We no longer needed dozens of tools; we just needed one device.
But something subtle happened in this transition. While we reduced physical complexity, we introduced cognitive complexity. Instead of carrying many tools, we started managing many applications. To complete even a simple task, we began moving across multiple apps, searching, comparing, messaging, paying, tracking. The burden of stitching everything together quietly shifted to us.
The Yellow Pages Never Really Left
In many ways, the app ecosystem started to resemble an older system we had already lived through: the Yellow Pages. When you needed a service, say a plumber, you opened a directory, browsed listings, compared options, and made a call. Each listing clearly stated what it offered and how to access it. The responsibility of discovery, evaluation, and coordination rested entirely with you.
Apps are not very different. They are simply digital listings of capabilities. The interface improved, the speed increased, but the underlying model remained the same: you search, you decide, you execute.
What is changing now is not just the technology. It is the role of the human in this system.
The Rise of AI as Orchestrator
Imagine building a house. You don't just "build a house." You find a builder, hire a plumber, coordinate an electrician, work with an interior designer, and manage timelines, costs, and dependencies. You are effectively the general contractor, orchestrating a network of specialised services.
This is exactly how we interact with apps today. Whether it is planning a trip, managing finances, or running a project, we are constantly coordinating multiple tools and services to get to an outcome.
Agentic AI changes this dynamic completely.
Instead of you acting as the coordinator, AI becomes the general contractor. You don't think in terms of tools anymore. You express an outcome: plan my weekend trip within €500, set up a business website, optimise my monthly expenses. The AI interprets the intent, discovers relevant services, evaluates options, and executes the workflow end-to-end.
In this world, apps don't disappear immediately. They become invisible. They turn into services, into Application Programming Interfaces (APIs), that AI can call on your behalf. The Yellow Pages still exists, but it is no longer read by humans. It is machine-readable, continuously evaluated, and dynamically used by AI systems.
The burden of navigation is removed. The burden of decision-making begins to shift.
From Orchestration to Absorption
What makes this transition more interesting is that it doesn't stop at orchestration.
In the physical world, when you build a house, you will always need a plumber and an electrician. Those are physical constraints. But in the digital world, many services are not bound by physical reality. They are informational or computational.
That changes everything.
If a capability can be generated rather than sourced, it no longer needs to exist as a separate service. This is where the second wave begins: absorption.
Some applications will remain as specialised services. Others will start to disappear into the AI layer itself. What used to require dedicated tools is increasingly becoming a capability of the model.
Content creation, basic design, simple coding, summarisation, planning: these are already being absorbed. The supply side is beginning to shrink.
The future is not simply one where AI connects to all apps. It is one where AI orchestrates many services and gradually replaces some of them.
A Shift in Where Effort Lives
If you step back, this evolution is not really about hardware or software. It is about where effort and responsibility reside.
In the physical era, effort was in managing objects. In the app era, effort moved into managing choices and workflows. In the AI era, effort shifts into defining intent.
This is a deeper transformation than it first appears. We are moving from managing the tools, to coordinating between them, to simply stating what we want. Each step is a compression of what the human needs to hold in their head.
What This Means
The implications are structural, and they work at every level of the industry.
First, apps become infrastructure. They don't vanish overnight, but they lose their visibility. What matters is no longer the interface, but the capability behind it. The last decade of competition on user experience starts to matter less. The next decade will be competed on whether your capability is accessible to an AI at all.
Second, APIs become the real products. If a service cannot be discovered, understood, and used by AI, it effectively becomes irrelevant. This inverts the logic of software design. You used to build for humans who would learn your interface. Now you build for AI systems that will call your endpoints without ever seeing your interface.
Third, AI becomes the primary interface. Not as a feature, not as a chatbot bolted onto an existing product, but as the control plane sitting between human intent and system execution. The question shifts from how well your app looks to how well your capability integrates into an AI-orchestrated workflow.
And finally, differentiation shifts. It is no longer about offering features or even better user interfaces. It is about delivering outcomes, faster, more accurately, and more personally than any alternative path to those outcomes.
Physical services (delivery, installation, healthcare, construction) cannot be absorbed into the AI layer, at least not any time soon. They will remain as services for AI to orchestrate. But every purely digital or informational capability is under pressure. If it can be generated, it will eventually be generated rather than sourced. The companies that survive this shift will be those whose capabilities are either physical-adjacent or genuinely beyond current generation quality.
Progress as Removal
We often think of progress as adding more: more features, more tools, more interfaces. But sometimes progress comes from removal.
The smartphone removed the need for physical tools. AI is now removing the need to navigate software.
We are moving toward a world where you don't open apps, browse menus, or compare options. You express what you want, and the system figures out how to make it happen. The apps don't disappear. They just stop being something you see. And eventually, for some categories, they stop being something that exists at all.
What is genuinely uncertain is the pace. The orchestration layer is being built now. The absorption of simpler capabilities is already underway. The displacement of more complex services will take longer. But the direction is clear enough that anyone building software in the next five years needs to ask a question that was irrelevant five years ago: am I building for a human to use, or am I building a capability for an AI to call?
The answer increasingly needs to be both. And soon, perhaps, just the second.
The Yellow Pages gave way to search engines. Search engines are giving way to intent engines. In each transition, the directory became more capable and the human became less involved in navigation. We are somewhere in the middle of the next one.
The apps are still there. For now, you can still see them.