In recent months, the major tech companies have been revealing distinct philosophies for how AI should be developed and experienced on their devices. A common pattern is emerging: AI is moving from the application layer into the operating system itself.
Historically, operating systems have acted as sophisticated switchboards. The bridge that connects users to applications, services, and information through UI that requires direct control.
We’re now in a position where AI is finding useful use cases in everyday life. Now we’ve all used AI, and know that sometimes it is great; sometimes not so much. The success of the product relies on certain conditions such as context, prompt quality, data quality, complexity, and human judgement. So as AI model owners look for as many places to implement their product as possible the natural home is an OS.
By embedding AI into the operating system itself, we move into a space where our devices can understand context, anticipate intent, and actively coordinate actions on behalf of the user. It can automatically satisfy a big chunk of the conditions needed to produce a good result; and the benefits are two way.
In this future, the role of the OS is no longer bound to managing hardware, applications, and interfaces. The OS can become a coordinated intelligence that orchestrates networks of agents, services, and data sources to create the experience for the individual user. The ability to reinvent, redesign, and reconfigure the more limiting switchboard based on what the user needs.
How this manifests across OS owners (Google, Apple, Huawei, etc) differs by philosophy, design and desired business model. But make no mistake, we’re moving from an intelligent operating system to a unified orchestration system.
The value for the individual
When we think of a truly valuable experience our minds often move into the service design processes that accompany luxury.
What makes luxury so sought after is the individual, bespoke, and concierge nature of it. Your needs are pre-emptively thought of, your problems proactively solved so you don’t encounter them. Decisions made on your behalf thought out to fit perfectly into your schedule, desires, and comfort. All whilst you remain in full control of the big choices, with the ability to completely change the direction in a single action. Everything is ultra-convenient. Everything feels incredibly human.
In the last ~20 years of mobile app development and digital experiences, our devices have followed a similar trajectory. We’re inundated with management apps – apps that require us to constantly manage the data within them. Calendars, contacts, messages, email etc. Expanding the definition loosely, even ride hailing, food delivery, and health apps are the same – entering data, filtering, organising, and comparing that data to make a decision based on what we want.
AI is being introduced in app experiences to reduce this cognitive load, and increase the value we get out of them. Embedding AI into the OS could remove the need to manage these apps entirely. In fact, it could allow your health app to talk to the ride hailing app, and the food delivery app. It will provide a luxury feel to data management.
Want to arrange time with friends? Your devices will hash it out to find the best time and suggest you meet for a morning run at the park in between you, since you’re both working towards the same health goals.
Want pizza? Based on your taste, here’s the one recommended option: a single click away, in your budget and it will be delivered by the time you get home based on your last meeting and commute.
It will redefine what good personalisation in tech looks like. It will redefine what convenience in tech looks like.
It’s ambitious and there’s a lot of technical questions that people smarter than I are answering in real time (I’m looking at you Nayan and the opportunity for the AI OS). Yet, based on the public announcements we’ve seen over the last month, it no longer sounds too out of reach.
What we’re hearing are two big questions that appear in this future, for the people trying to create value on these OS’.
- How do I make sure I am the pizza being recommended?
- If an agentic OS is going to be the one ordering the pizza, what capabilities do I need to build into the app I build?
Being the chosen pizza
We should expect significant changes to how customers discover and interact with digital services. Many transactional interactions that currently begin with opening an application may instead begin with an AI assistant or OS level agent.
Your discoverability optimisation for these agentic experiences should focus on being the service the agent chooses. That’s either through rethinking digital products accessibility, interoperability, trust and machine-readable value (design and tech working hand in hand). Or through convincing the customer to request you as the service of choice from the get go.
A familiar comparison is having your product featured in a supermarket. Traditionally launching a digital product means getting your product on the supermarket shelf and convincing the customer to buy you out of all the options. The difference with AI, you’re convincing the store clerk to recommend your product, so when a customer states their goal, problem, or desire, they take them directly to you and bypass the other products entirely. Or having the customer ask the clerk specifically for your product in the first place.
As of this writing, the best approach we can see for AI discoverability is still GEO, which we have discussed here. Human discoverability – well that’s brand building and trust.
The bigger conundrum
The longer-term implications of the app layer remain uncertain.
If the OS becomes capable of understanding intent, orchestrating services, and completing tasks autonomously, the role of traditional apps may fundamentally change.
In other words, how much of the user journey is going to collapse into the OS? How will our commercial models be affected if the agent comes knocking for a finder’s fee? Will the OS itself become the primary destination for digital interaction; absorbing a curated journey inside of a singular app, let alone the thousands of separate products and experiences on the app store? Do I need to build an app and do I need to do it fast?!
Building a human focused app
“Mr Watson - Come here - I want to see you”
It is poetic that the phrase first said over the telephone, as recorded in Alexander Bell’s notebook in 1876 to his assistant, was the organisation of a conversation to be had off the device. You just brought to life a technology that allows you to communicate vast distances in real time, but screw that, let’s chat face to face.
It can be easy to lose track of what we are trying to streamline, optimise, and scale – that the device is there to support our lives, not become them. When we start to consider how information is processed, and delivered to us through our devices, and how we want that information to be delivered; we have to have a deep understanding of why someone engages with that information in the first place.
In our work designing and building next-generation AI-enabled services, we’ve found that the true value of apps doesn’t come just from the data it provides, but the human change they enable and the joy they create as they do it.
We are deeply complex creatures, and we’re motivated not through a lack of knowledge, but the emotion that we care about making change.
In the era of the AIOS, your app dies when the human stops caring about which app does the job
We’re seeing this play out in real time in the creative content space. AI tools enable anyone to create content fast, cheap, and scalably. This has led to a lot of slop, which society is reacting to. It’s a contradiction – AI has made content easier than ever to create, but harder than ever to make stand out.
Creators are responding to this change, by adapting their content, increasing the tangible production value, and showing behind the scenes. Cheap, quick content is no longer viable, because people don’t engage with it in the same way.
The same is true of digital products – more opportunity to create but harder to make commercially viable. The AIOS will only amplify this by raising the quality threshold required to earn someone's attention.
Making an app that people stick with is about demonstrating that you can deliver the outcome someone is looking for unlike anyone else. It’s having a distinct and valuable role in someone’s developing sense of self, and their holistic digital services they engage with. Always grounded in human emotive outcomes that economically, might not make sense to develop – but positively reinforces moments of joy.
This opens up a playground for us to explore unlike anything before. The floor has been raised. A workout app that never needs opening will become standard – we have to be delivering something that goes beyond that. Something that demands to be opened not because it doesn’t work otherwise, but because it is the best way of experiencing it.
That requires more work.
The theory is that in the AIOS that can deliver outcomes to the user without ever needing to open an app – the apps that remain valuable are ones that go beyond transactions.
Knowing where to focus today to prepare for the future isn’t always obvious, so if you're navigating these questions, let's talk.