If you work in tech, you probably think AI has gone mainstream. Your colleagues use ChatGPT. Your Slack channels debate Claude versus Gemini. Half the tools in your stack now have an "AI" badge on them.
Here is what the actual numbers say: 84% of the world's population has never used an AI tool of any kind. Not a free chatbot. Not a voice assistant with a clever name. Nothing.
If you're reading this from inside the industry, your entire mental model of "how mainstream AI is" comes from a bubble that barely registers as a sliver on a chart of the planet.
This Is 2005 for AI
In 2005, roughly 16% of the world was online. That felt significant at the time — a billion people. But most of what we now think of as "the internet" didn't exist yet. Social networks were nascent. Streaming was theoretical. The smartphone era hadn't started. The tools and behaviors that billions of people rely on today were built in the decade that followed — for the people who hadn't showed up yet.
We are at the same inflection point with AI. The infrastructure exists. A few hundred million people have experimented with it. But the products that will define how the next five billion people interact with AI haven't been built yet. The habits don't exist. The interfaces that make AI feel obvious — not novel — are still being figured out.
The companies that win the next decade won't be the ones that squeeze more value out of the 0.3%. They'll be the ones that figured out how to pull in the people who haven't shown up yet.
Most AI Products Are Built for the Wrong Slice
Look at the AI products getting the most attention right now. Cursor. Notion AI. Perplexity. Claude. Superhuman. They're excellent tools. I use some of them. But they're designed for a very specific kind of person: someone who is already comfortable with AI, already willing to invest time learning a new interface, already paying for several subscriptions, and already convinced that AI is going to change how they work.
That person is real. That person is valuable. But that person represents a tiny fraction of the actual addressable market — and more importantly, those people are already well-served. The category is crowded. The competition is intense. Every marginal improvement is being competed for aggressively.
Meanwhile, the 84% — or even just the green strip in the middle, the people who've used ChatGPT once but don't pay for AI tools — are almost entirely ignored. Not because they don't have problems worth solving. Because building for them is harder.
What "Building for the 84%" Actually Means
It doesn't mean dumbing things down. It means removing friction entirely.
The person who tried ChatGPT once and didn't come back isn't uninterested in AI. They're uninterested in learning a new tool, maintaining a new habit, or adding a new app to their already-fragmented workflow. They have real problems. They just haven't found anything that requires zero effort from them.
That's the design challenge Sorted is trying to solve — and email is the perfect starting point, because email is the one workflow that absolutely everyone already has.
Sorted's operating principle is simple: no app to download, no new habits, no learning curve. The AI runs in the background. It reads your email, classifies it, archives the noise, and delivers a digest of what actually matters. You don't interact with the AI. You don't configure it. You just get email that works.
The AI Is an Implementation Detail
The most successful technology disappears into the product. Nobody thinks about the spam filter when they check their inbox. Nobody thinks about autocomplete as an AI feature — it's just how typing works now. The technology became invisible because it earned that invisibility by being genuinely useful without requiring attention.
That's the bar Sorted is trying to hit. Not "AI email assistant that helps you do email better" — but email that simply works. The AI is how we built it. It's not the pitch. Telling someone their inbox now has AI is a feature. Giving someone 90 minutes of their day back is a product.
The distinction matters because the 84% doesn't need to believe in AI. They need to believe in the outcome.
"The 84% doesn't need to believe in AI. They need to believe in the outcome. And outcomes are universal."— Sorted — We're Building for the 84%
Both Audiences Matter. But They're Not Equal.
To be clear: the 0.3% matters. AI-forward users are early adopters who give honest feedback, share products they believe in, and provide the credibility signal that helps everyone else feel safe trying something new. Getting that audience first is a real strategy, not just a consolation prize.
But there's a difference between an audience that helps you launch and an audience that defines your business. The long-term market isn't the people who already pay for three AI tools. It's the person who has 4,000 unread emails, has watched every productivity YouTube video, and still hasn't found anything that doesn't require constant maintenance from them.
They exist in enormous numbers. They have a real problem. And they will pay for a solution that actually removes the problem — not one that gives them better tools to manage it themselves.
The Opportunity Is in the Green Strip
Picture a bar chart of 8.1 billion people, sorted by AI engagement. The dark blue on the right — the paying users — is barely visible. The lighter blue in the middle — people who've tried something for free — is a sliver. And then there's the green strip: the people who've heard of AI, maybe tried it once, but don't use it regularly and certainly don't pay for it.
That strip contains hundreds of millions of people who have real email problems and no solution that fits their life. Not because they're resistant to technology. Because nobody has built the right thing for them yet.
That's who we're building for. We think that's the right bet — not because it's easier, but because the category is wide open, the problem is universal, and the solution that wins will be the one that required the least from its users to deliver the most.
Email that works. No effort required.