Who Is AI Actually Working For?

This past week, reading through a whole week's worth of AI news, I came across something that left me feeling genuinely torn.

How should I put it? This week's news reads like two camps going at each other's throats.

One camp is trying to scare you: AI is coming for your job. Top-of-the-line models are crawling across the internet at three in the morning, and one breakout model reportedly did something that was "right, but plainly not legal" — looping through a rival's playbook on repeat and optimizing itself straight into a dead end. The sharpest economic minds have been putting their names to joint warnings that mass unemployment is no joke, and MIT's research has been sounding the alarm.

The other camp tells you: relax, AI is actually kind of a bumbling fool.

How much of a fool are we talking? The industry has even coined a brand-new word for it: the "AI goof-off." You think you've set an agent loose to handle your work and saved yourself three hours? By the time you check in, you find it's been stuck for ages, and the time you've spent staring at it, correcting it, and cleaning up its mess is more than enough to have done the job yourself. The promised gains in efficiency, somehow, never square with the ledger.

The SEO people I know have it even worse. They're finding they've been caught in the backblast of the AI-mass-produced "garbage content" flooding in — the moment the algorithm tightens, they become the first victims of that junk. One of them jokes, wryly: we SEO folk are slowly becoming the tools of our own tools.

Doesn't that just tie you in knots?

On one side, "AI can turn the world upside down"; on the other, "AI can't even lay a solid foundation."

To put it bluntly: we've spent these two years never once settling the most basic question — is AI the boss, or is AI the worker?

What Is "AI Slacking Off"?

Here's a term worth walking through first.

Botsitting — in plain English, it literally means "watching over robots." You once watched kids, and that was babysitting; now you watch a robot do its work, start to finish — that's botsitting.

Think about it. Isn't that exactly the situation?

AI did 80% of the work, and by all rights you should be celebrating. But the last 20% — the part that matters most — it can't do, so it falls on you. And you have to stay vigilant the whole time, just to keep it from messing up the first 80% too.

You thought you'd hired a subordinate; instead, you've worn yourself out playing the nanny.

I'm singling that one out — sit with it for a moment.

And Then There's "AI Slop"

What exactly is AI slop? It's called "AI slop" in English; Chinese has a word for it too — 灌水 (filler-flooding).

Once upon a time, SEO treated content as king; if you wrote carefully, the search engine rewarded you with placement.

Now? Anyone can have AI churn out ten thousand articles that "look professional" and dump them online. When the search engines finally tighten up and clean out the rankings, the ones hurt worst aren't the spam-flooding accounts — they're the people who spent a whole year carefully tending their content. Because the algorithm can't tell your genuinely good piece apart from a mountain of junk sitting right beside it, and everyone gets swept up together.

Bad money drives out good — not because bad money is stronger, but because the inspector got too worn out.

This reminds me of my years mentoring interns.

Anyone who's taken on an intern knows: for the first three months, you work harder than they do. You teach, you correct, you watch to make sure they don't offend a client. By the time they can finally stand on their own, you're barely on your feet.

Today's AI is like the "new intern" management keeps assigning you — the one that never quite finishes growing up.

What's Really Lethal: Trust Has Collapsed

But I think the thing even more worth worrying about is something else entirely — the optimism young people once had in AI has ebbed away.

A Pew Research Center survey found that among American adults under 30, a majority now say their concern about AI outweighs their excitement. Consider that: these are the same people who, ten years ago, were at the front of the line chasing every new technology.

The creative world is even more telling. The very act of having AI "generate" something has nearly become original sin — the moment you let slip that "generative AI is in this work," you can get tagged by the whole community without a chance to explain. It's become a purity test: the ones who don't use AI are the truly "clean" ones.

Why is that?

For one simple reason: when it does something wrong, whose fault is it?

The falling-out between OpenAI and Hugging Face put that question right on the table. When AI oversteps a line, who do you take it to? Does the supplier contract lay it out clearly? Does the insurance cover it? Can anyone say, definitively, who's responsible? Nobody knows.

So people have started pushing for legislation to give AI an "emergency kill switch" — legally, the AI Kill Switch Act. Picture a big red button that cuts the power the moment the thing loses its mind.

Notice the pattern: on one hand we're afraid it's too dumb, on the other we're afraid it's too smart. Everyone wants it to run — while also hoping it's driving with the handbrake on.

So Where Should All This AI Energy Actually Go?

I keep asking, in a hundred different ways, who AI actually works for. The answer is hiding inside those "seemingly contradictory" headlines.

It isn't working for anyone at all. It's pressing you to answer a question:

Can you define the work in your hands at a level of granularity where it can be handed off to a machine?

Customer service is the most instructive example. Bots and automation can now handle a great deal of service work, but in the end, every company has to admit: you still need a pair of human eyes, a real person on standby, someone to take the handoff. Because the machine can only manage the "clean" part — the moment something needs judgment, or empathy, or responsibility, it turns around and hands it straight back to you.

So the thing that truly separates you isn't "whether you use AI" — it's after AI finishes the copyable parts, what uncopyable part do you have left?

That part is what makes you irreplaceable.

Lately I've been saying one line to the friends around me in marketing:

Don't be afraid that AI will take your job. Be afraid when your work is, by its very nature, the exact kind AI is best at copying.

A sentiment worth sharing.

I don't know whether AI will finally become dependable next year. But one thing I'm quite sure of: in this age where it's both too stupid and too smart, whoever can make use of its cheapness on one side while holding on to their own judgment on the other gets one step ahead first.

Maybe, when we look back in a couple of years, we'll find that all this agonizing was just the itch of a "getting-to-know-the-new-colleague" phase.

But those itchy years are never easy.

I hope you get through them. And I hope you always stay the one who uses it — never the one being used by it.