The end of apps? Not so fast.
The traditional computing “stack” is usually portrayed as a layered cake where the base is silicon, on top of which are system-level interfaces, followed by the operating system, then core services, and finally applications. The User interacts with the applications only. This is an app-based ecosystem which exposes “tools” for users to use as they try to accomplish tasks. This ecosystem for Apple is approaching $2 Trillion/yr in economic transactional value.
The diagram above, introduced as the new Apple Intelligence architecture, is quite different. The user is connecting using various devices via various input methods (image, voice, text) to prompt an orchestrator to tap various knowledge bases (local and remote) to initiate either Siri or an App to act on some intent.
The role of the operating system is now that of “orchestrator’ that connects inputs to knowledge and decides how best to fulfill an intent. It sometimes connects to a chatbot (Siri) but also initiates apps to act.
This might be seen as moving from an app based ecosystem to an intent centric world. For the original Siri, queries were usually routed to apps (calendar, weather, stopwatch, music, etc.) In this new world Siri AI will resolve the intent directly and have the ability to construct & deconstruct interfaces / experiences on demand. There are many tasks that do not need third party applications.
However, the diagram still shows Apps as equal to Siri in terms of Systemwide Experiences. Apple announced a Foundation Models framework and expanded App Intents. Intent-centrism doesn’t deprecate apps. It transforms them them from intent destinations to callable functions.
Using the nomenclature of integration/modularization and business model evolution, it modularizes what was integrated, subtly shifting the locus of value. The App Store becomes an API directory. That’s a business model shift. Classic innovation theory suggests that as the value chain evolves, some profit will evaporate from one point and condense in another point in the chain; usually wherever the new point of modularization has been injected. Subscribe to Asymco One for full access to our research.
The old points of concentrations were the OS (where Apple captured profit) and the Apps (where developers did.) The new points might be in the orchestrator and in the foundation models.
But I don’t think it’s that simple.
Models might be commodities and orchestrators are not much different than OSs. Intent action requires access to private app data. bank accounts, calendars, messages. Authentication is an app responsibility and thus earned trust. Every app removed or disintermediated becomes a permission the orchestrator needs. Would banking credentials be granted directly to the device/OS? Much peril lurks there. This local data is where app control lies.
Where the threat for apps lies is in the possible loss of user interaction. It’s also a great part of the cost of development. Shifting from a point of contact with the customer to a callable function comes with costs and benefits that are not easily predictable.
Transitioning the OS from a layer between hardware and applications to a layer between foundation models and experiences fundamentally shifts the architecture, but, in the case of Apple, it preserves its value as the agent of trust and discovery.
For this the value created and captured remains and perhaps increases. It’s clear to me that Apple has engineered this quite carefully and thoroughly.