Piero Savastano
The Graphical Interface Is Dead

The Graphical Interface Is Dead

October 31, 2026
5 min read
Table of Contents
index

Now, I should really keep my mouth shut, since I lost my job because of artificial intelligence and now I’m on unemployment benefits. But I want to say this: all these people building applications right now, all the software houses making chatbots, want to compete in the market of years ago with today’s tools, without taking into account that anyone can do it now.

The entry point is the Big Tech chatbot

I invite you to think about this. Hundreds of millions of people are starting to use ChatGPT and Claude, and these things are built precisely to navigate data and services in place of the people themselves.

Example: I want to organize a day trip to the countryside. You tell Claude directly: I want to go there, more or less at this time, these are my constraints, show me something, give me a few proposals. Then maybe you go and browse some websites, but the entry point is the chatbot. And it’s not the chatbot of Joe Bloggs Ltd. It’s the chatbot of Big Tech.

So building chatbots and applications, these graphical interfaces - admirable as it is that someone sits down, studies, and pours their experience into it - is becoming less and less relevant. My invitation is to focus on the data service, and to stop thinking of your application as a landing - the place where people arrive. Because as time goes on, people stay more and more inside the agent of Big Tech, not yours.

This is a reflection I often bring up in my corporate courses, Reskill - get in touch if you’re interested. Again, corporate courses only, not individuals. For individuals I’ll do something directly online, don’t worry.

Agent as browser, or agent as service?

The question I often ask is: does this agentic economy see the agent as a browser, or the agent as a service? What do I mean?

The agent as a browser means the agent - whichever Claude of the moment - is close to you, and you use it instead of Chrome. You talk to an agent, and the agent navigates the data services. Going back to the trip: the agent helps me define the constraints, figure out roughly what we’re talking about, and then maybe I jump somewhere else, but most of the browsing is done by the agent, not by me in the browser.

The agent as a service means there are various agents talking to each other, and every endpoint, every online service, has its own agent that I have to go to. To talk to the Booking agent I have to go to Booking, to talk to the email agent I have to go to my email.

In my opinion, it’s not like that. In my opinion the agent is the entity that browses, and so it’s the one close to you - at least in this historical period.

How do you make room for yourself?

So how do you carve out space in an economy like this, for the near future at least, since who knows what happens after? Certainly not by building yet another chatbot or yet another application, with yet another landing page, yet another login, yet another set of screens.

I don’t mean any disrespect to anyone, because everything is fragmented and nobody understands a thing. But my bet is that this stuff is no longer useful at all. The very idea of a web graphical interface designed entirely to carry out a certain function is already old, already obsolete.

You don’t need to welcome people on your graphical interface. You still have to fulfill a request, but you fulfill it by serving an agent, which in turn is in contact with the user. Especially for us small players - the long tail of services - we are no longer in contact with the user. We are in contact with the agent. So the data service matters much more than the graphical interface.

And if you have visual experiences, things to show, you don’t show them on your own website. You show them somewhere else - which I won’t tell you here. Reskill, get in touch.