Why I think we'll use far fewer apps
Software isn't going away. You just won't be the one opening most of it.
12 October 2026
Recently I decided to cancel SiteGround, which meant moving the DNS for one of my domains to Namecheap without breaking the website or the email that runs through it. That used to be an evening of logging into two dashboards, copying records across by hand and hoping I hadn’t missed one.
I said it in a chat. My AI OS took a snapshot of the old records, moved the nameservers, kept the site and Google mail pointing where they should, added the new subdomain I needed and checked it all resolved. I didn’t open SiteGround or Namecheap once.
That’s most of my week now. Updating a website, checking whether an invoice has been paid, finding what someone said on a call last month: I ask, usually by voice, and it goes and deals with whatever software it needs. Often I couldn’t tell you which dashboard it logged into.
I don’t think that’s a quirk of how I work. I think it’s where software is going. There’ll be more of it than ever, but most of the time you won’t be the one opening it.
Today
You open each one, find the right screen and do it yourself.
Soon
You say what you want. It deals with whatever software it needs.
We already have too many
Zylo’s 2026 SaaS Management Index found the average company now manages 305 SaaS applications, and leaves 36% of its licences unused. On our phones, Sensor Tower puts the time we spend in apps at about 3.6 hours a day per person.
Each of those apps knows a thin slice of your life and very little about the rest. You’re the integration layer, carrying information from one to the next.
Most apps are a database with some screens on top
Strip most business software back and it’s a database, some logic and a lot of screens for getting things in and out. The screens are what you pay for and where you spend your time. But they only exist because a person needed some way to use it.
If an agent is doing the using, most of those screens are in the way.
Dashboards are the clearest example. Every app has its own, and each one only shows its own slice. I still like dashboards, but I’d much rather have one, pulling from everything I use and built around what I actually care about, than twenty I have to go and check.
Your AI becomes the front door
The interface I want is a conversation with something that knows me well: what I’m working on, who I’m working with, and what I’ve already agreed. It can act across all my services, which is a much better experience than switching between twenty apps that don’t talk to each other.
It’s a bit like having a really good PA, except it’s available to everyone, not just people running large companies. It also makes a lot of the friction we’ve put up with for years look ridiculous, like typing the same details into five different forms.
Plenty of people are already heading that way. Pew found that half of American adults now use AI chatbots, and looking things up is the most common reason, ahead of work.
Notifications go the same way. Almost every app is fighting for your attention, because that’s how it’s measured, and your phone becomes a constant stream of things that mostly don’t matter. Your AI becomes the filter and the one place things come from. It deals with a lot of it without you ever hearing about it, and only interrupts you when it’s worth it.
Businesses will have to work for agents
This is the part I think most companies are underestimating. If more of your customers send an agent instead of visiting themselves, your website, your booking system and your support all have to work for that agent.
It’s already started. Adobe found that traffic to US retail sites from generative AI tools rose 693% over the 2025 Christmas season compared with the year before. The numbers are still small, but they’re moving fast.
In September 2025 OpenAI launched Instant Checkout, so people in the US can buy from Etsy sellers without leaving ChatGPT, with over a million Shopify merchants to follow. Google, Visa and Mastercard have all published their own ways for agents to pay. The plumbing for agents doing business on your behalf is already being laid.
If an AI can’t easily find out what you do, check your availability and book something, it’ll go somewhere it can. A decent API, or an MCP server, is going to matter more than a beautiful app. Anthropic counts more than 10,000 public MCP servers already, and handed the standard to the Linux Foundation in December 2025, with OpenAI as a co-founder.
Software for one
The other shift is that a lot of software will be built for one person or one small team. When building something costs almost nothing, it stops making sense to bend yourself around a product designed for thousands of other people. You describe what you need, have it built, and throw it away when it stops being useful.
I’ve built dozens of small tools that only I will ever use. Some of them lived for a week. A few years ago that would have been absurd. Now it’s normal.
It isn’t that everyone becomes a developer. It’s that the line between using software and making it gets blurry. Asking your AI for a tool that does something and asking it to just do the thing start to feel like the same request.
The catch is trust
None of this works without trust. You’re giving one system access to your email, your calendar, your files and eventually your bank. I understand why people are nervous. For me it’s worth it, but only if it’s set up properly: sensible permissions, and anything consequential coming back to you for a yes before it happens. I’d also much rather that system was something I control than something one big company owns outright.
Who owns the front door
That’s the real question for the next couple of years. The AI labs and the phone makers are the obvious candidates, because they already have the model or the device. ChatGPT alone had 900 million weekly users by February this year.
But you don’t have to wait for them. You can set up your own version today, connected to your own stuff and built around how you work. That’s what I’ve done, and it’s why I’m fairly confident about all of this.
It isn’t a prediction about a distant future. It’s a slightly more extreme version of how I already work.
Sources: Zylo, 2026 SaaS Management Index · Sensor Tower, State of Mobile 2026 · Pew Research Center, Americans and AI (Jun 2026) · Adobe, 2025 holiday shopping season (Jan 2026) · OpenAI, Instant Checkout (Sep 2025) · Anthropic, Donating MCP to the Agentic AI Foundation (Dec 2025).
I co-founded EmbedAI and I'm a co-founder and Head of AI at FanFinders. I write about running my life and work on an AI OS.
Get the next one
New pieces like this, and first word on the book.