Tech leaders are trotting out the “We’ll use AI to reduce headcount in the next 2 years” talking point again.
I’ve been hearing this talking point for at least two years now. It’s always just a couple years off. It’s not backed by the actual ability for AI to do that, of course, but it sounds good to investors and it’s a good way to shut up those pesky workers with their demands for better treatment.
We think AIs are transformative because they can do things we’ve never seen a computer do before, but they have the giant catch that you can’t really rely on their output to be correct and deterministic the way you could with simpler computer algorithms. They’re just another piece of new tech delivering incremental improvement with tradeoffs.
I’m not suggesting AI tools have made no improvements recently. But we’re collectively pouring massive capital into evolving these models and we’re getting diminishing returns. We’re not going to get to a point where they “just work.” It’s literally not what they’re designed to be able to do, and new features like agent modes and MCP servers are still papering over that really important detail that LLMs are still just generating statistically plausible output given a particular input, without regard for understanding what’s underneath.
But suppose for the sake of a hypothetical that I’m wrong.
Let’s assume AIs actually are about to become so accurate that you can feed them a prompt and they’ll just produces the output exactly as you expect, every time.
It doesn’t have to play out like some dystopian scenario where everyone gets fired and the world falls apart.
Fully Automated Luxury Gay Space Communism Utopia
Picture it: the AI that just works has landed.
While all the biggest companies try to one-up each other on who can fire as many people as possible while doing as little as possible for their customers with the new tech, the rest of us get to work creating a better world.
Instead of just using the AI to fire staff and rake in more profits, we could provide substantially better products and service for customers.
The bugs that used to lay neglected for years in the icebox get fixed because the AI helps fix bugs almost as fast as you can report them. Those feature requests that a vocal minority of customers clamored for but never got? We can implement them now and those customers rejoice. Eventually the backlog is empty. When a customer reports a bug or requests a feature they might see something ship in a week or two.
Getting support is a breeze. The AI-powered chatbot actually is helpful, and there are plenty of humans left over so that when there is something the AI can’t handle, they can step right in and help. And they’ll do an even better job because we’ll have good internal tools for those humans to use.
And while we’re at it, we switch to a four day work week, and work fewer hours a day. The AI is helping out so much that it’s stupid not to reduce hours. People are less stressed. They stay in jobs longer. They take longer vacations.
And because it’s so easy to spin up a new business now, even starting out solo, a lot more people do just that, and Big Tech’s stranglehold on the world gets chipped away at from every angle. Big Tech doesn’t just get outmaneuvered by more agile companies, they’re getting out-executed too. Most of the structural advantages you had as a big company vanished.
I’m so sick of the same uncreative takes from people who seem to be getting off to the idea that we are on an inevitable path to dystopia.
Again, AI is not going to get that good. But if we did, we could make something better than a dystopia.
And frankly, we could use the tools we have now to work toward making something better than a dystopia.
Leave a Reply