The Death of the App: What Comes After Screens, Menus and Buttons?

For twenty years, the answer to almost every digital problem has been: “There’s an app for that.” Need a taxi, a thermostat setting, a parking ticket, a blood-pressure reading, a sandwich that has travelled three miles in a padded bag? App.

It was a sensible answer for a while. Smartphones put a capable screen, network connection and reasonable processor in our pockets. Apps gave services a neat, branded little home. Then every organisation built one, whether it had a coherent reason to or not. I have seen temperature-monitoring systems that could tell you a freezer had failed at 3am, yet required six taps and a password reset to acknowledge the alert. Progress has occasionally had the manners of a wet dog in a small kitchen.

The app is not about to vanish. But the idea that every interaction needs a separate screen, menu hierarchy and button-filled miniature website is beginning to look rather dated.

The app has become the problem

Modern phones are full of software most of us use once, reluctantly, and then forget until it demands an update while we are trying to pay for something. The average organisation treats its app icon as if it were a tiny embassy on your home screen. The user experiences it more like administrative clutter.

The fundamental problem is context. An app asks us to stop what we are doing, find the right icon, authenticate ourselves, navigate its internal logic, and translate our real-world need into the categories its designers imagined months earlier. Want to know whether the medicine fridge is safe? You may have to open an app called something cheerfully opaque, select a site, select a cabinet, select a date range, and finally discover the answer is “probably”.

That model is particularly poor for people who experience fatigue, reduced dexterity, cognitive overload, visual impairment or simply the ordinary human condition of being busy. As a disabled engineer, I am painfully aware that an extra step is not always trivial. On a bad day, a login screen with fiddly controls can be the difference between doing something independently and deciding it can wait. Technology loves to call this “friction”. Users usually call it annoying.

From destinations to capabilities

What comes next is less about replacing apps with a single magical interface and more about making services available where they are needed. Instead of going to an app, we ask for an outcome.

A good assistant might let you say or type: “Has the vaccine fridge stayed between 2°C and 8°C overnight?” It should return the answer, show the relevant evidence, and flag an excursion if there was one. It should not respond by opening a dashboard and wishing you luck.

The same principle applies at home. “Turn the heating down when I leave.” “Pay this invoice on Friday unless the amount changes.” “Tell my sister I will be late, but do not make it sound dramatic.” These are intents, not app journeys.

AI is making this more plausible because language can finally become a usable input method rather than a parlour trick. But language alone is not enough. Anyone who has tried voice control while a television is on, a kettle is boiling and a cat has decided to contribute knows it is not a universal interface. Voice is useful. So are text, cameras, wearables, physical controls and automation. The future is multimodal, which is a rather grand word for “use the method that works right now”.

Ambient systems need boundaries

The appealing version of this future is calm technology: systems notice routine events, do useful work, and ask only when a decision genuinely needs us. My monitoring work has always been heading in that direction. A sensor should collect readings quietly. Rules should identify a genuine problem. A human should receive an alert that explains what happened and what action matters. Nobody needs a notification every time a perfectly healthy fridge remains cold. That is just a machine seeking praise.

However, there is a serious trap here. Moving beyond screens can also move decisions out of sight. An assistant that books appointments, spends money, changes access permissions or interprets health data needs clear limits, audit trails and an obvious way to correct it. “The AI thought it was helpful” is not an acceptable incident report, even if it has a pleasingly modern ring to it.

This is where much of the current AI excitement becomes rather slippery. A conversational interface can hide a poor system behind charming prose. If the data is stale, the permissions are vague, or the workflow is unsafe, an assistant merely makes failure more chatty. The boring engineering still matters: reliable data, explicit state, identity, security, fallback paths and testing with actual humans.

Accessibility should shape the next interface

There is an opportunity here to build something better than the app era, but only if accessibility is treated as a design input rather than a compliance afterthought. Systems should retain preferences without making users repeat themselves. They should offer short routes for frequent tasks, plain explanations for important decisions, and meaningful controls when automation gets it wrong.

That means designing for variability. Some days I can happily investigate a graph, configure an alert rule and argue with a wireless gateway. Other days, I want a system to tell me plainly what needs doing and leave the decorative analytics for later. Both are legitimate ways to use technology.

Screens, menus and buttons will remain, just as paper forms survived the arrival of databases. They are excellent when we need to inspect detail, compare options or retain control. They are dreadful when they become ritual obstacles between a person and a simple outcome. The next generation of products will not kill the app with a flourish. It will quietly make many app-shaped tasks unnecessary, which is probably the most useful kind of revolution.