AI replaces conversation — but not for the rich

According to Common Sense Media, 72% of American teens have tried AI companions, and a third have chosen a bot over a human for a serious conversation at least once.

Author: Michael Kokin ·

Live human connection seems to be becoming a luxury good. Robb Report ran a story about how AI is changing premium service. Here's my retelling, plus my own thoughts.

London restaurateur Jeremy King told his staff he could replace them with robots. They don't show up late, they don't forget anything, and they don't come in hungover. But he's not going to: robots have no empathy, no intuition, no capacity to love.

For twenty years, Silicon Valley has promised to strip away every unnecessary interaction. No need to talk to a waiter — order by QR code. No need to talk to a salesperson — pick by algorithm. No need to write — AI will do it for you. And here's the paradox. At an expensive restaurant, part of what we're paying for is a waiter who remembers our table, and three fittings with a tailor are part of the service. If a restaurant's only job is to feed you, the best one should serve you the fastest. But that's clearly not the only reason we go.

Rory Sutherland of Ogilvy uses the doorman as an example. If his job is just opening a door, an automatic one would do fine. But a doorman recognizes guests, calls cabs, watches for security. With him around, you feel like you're expected here. So by optimizing away the function, we lose something we can't put a number on.

Journalist Kathryn Jezer-Morton coined the term "friction-maxxing" — deliberately giving up certain conveniences for the sake of human experience. Talking to a stranger instead of reaching for your phone one more time. This isn't just nostalgia: research by Sandstrom and Dunn links even casual contact with neighbors and acquaintances to a sense of well-being.

What the numbers say

According to an Elon University survey, 27% of adult US internet users talk to AI about personal and emotional topics. Common Sense Media found that 72% of American teens have tried AI companions, and a third of users have chosen a bot over a human for a serious conversation at least once. According to US time-use statistics, Americans now spend roughly a quarter less time socializing than they did in 2003. Though the decline started long before ChatGPT, so AI's role here isn't proven. We covered a similar pattern in why people turn to AI for health advice.

MIT professor Sherry Turkle notes that wealth now chases scarcity. The rich used to dream of getting rid of staff they had to talk to. Now that everyone has an AI assistant, they want people back — people who actually give a damn.

Why we need other people

I work in AI, I build autonomous agentic systems, and I think automation is a massive achievement. I like the idea of a world without pointless work. But I keep coming back to the same question: what are we actually going to do with the time we free up, and why do we even need other people if they're no longer required to get our problems solved?

This is one of the central themes of my OSY project — about the meaning of life after automation. We're used to measuring people by their usefulness. An employee is productive, a consultant solves a problem, a friend helps out. But what happens when AI learns to do all of that faster and better? We'll have to rediscover the value of relationships that don't have to be useful.

What worries me

That said, I'm not romanticizing human interaction. A robot can be a better waiter, and AI can be a more attentive conversationalist. There's no need to glorify long lines and bad service. What actually worries me is different: we're automating away not just the unpleasant chores, but the need to deal with other people's will, unpredictability, and imperfection. And that might be exactly where human experience begins.

If genuine human attention becomes a premium service, we risk ending up with a society where the wealthy get actual people — teachers, doctors, conversation partners — while everyone else gets a quality imitation of engagement. Technologically, that might be more efficient. Socially, it would be a pretty strange kind of victory.