AI Isn't Helping You Work. It's Replacing Your Decisions.

A while back, I went through reports from a dozen or so media outlets from the first half of 2026 — all about AI and marketing, AI and the workplace, AI and companies.

When I finished, a chill ran down my spine.

Not because AI had gotten more powerful again. It was because I realized that nearly every outlet was telling the same story: everyone pays lip service to "using AI to boost efficiency," but their actions tell a very different story — they're using AI to cut jobs.

Those two things are worlds apart.

Boosting efficiency means the work that used to take ten people still takes ten people — but each person's output doubles.

Cutting jobs means the work that used to take ten people now takes three people plus a few AI tools, and the other seven are shown the door.

You see — one is "helping you," the other is "replacing you." Two completely opposite directions.

A Hidden Food Chain

Let me first walk you through what's actually been happening across the media landscape these past six months.

Webflow, a company that builds website tools, one day suddenly revoked system access for large numbers of employees — their keycards wouldn't even open the doors. People didn't even have time to say goodbye before their desks were cleared out. The report used a word: "bloodbath."

This wasn't an isolated case. Company after company discovered something that thrilled them: when a position opens up, you don't necessarily have to refill it. Pile the work onto the people who stay, add a few AI bots, and somehow the machine keeps running.

Economists started sounding the alarm: this could be a trap. Once a company discovers that "fewer people plus AI" gets the job done, they won't go back to hiring. The blade of layoffs doesn't fall — it slowly tightens. You don't even feel the pain, until one day you realize an entire department's headcount has quietly shrunk by half.

That's the first food chain: companies are using AI to redefine "how many people a role actually needs."

The Word "Collaboration" Is Lying to You

There's another buzzword all over the media — AI agent.

A lot of vendors love to package it as "your digital colleague." Sounds so friendly — like having a new buddy at the next desk who never complains and never takes a sick day.

But an MIT Technology Review article ripped off that fig leaf with its headline alone: AI agents are not your colleagues.

Why?

Because there's one thing between colleagues that AI simply cannot provide: pushback and mutual accountability. When people work together, they naturally call each other out, cover for each other, and catch each other's mistakes. You slack off, your partner holds you accountable. You make a careless error, your colleague snaps you out of it. This "imperfect but mutually tethered" relationship is precisely the source of organizational resilience.

AI won't snap you out of anything. AI will just follow your instructions and charge full speed ahead.

TechTarget compiled a list: four "resilience incidents" caused by blind trust in AI. Every single time, someone treated AI's output as fact — no one double-checked, no one hit the brakes, and the entire organization ended up getting bitten by the very errors it had fed itself.

So stop letting the word "collaboration" fool you. AI is a tool, and a tool's job is to be mastered — not to be treated as a colleague. The moment you treat it as a colleague, you've handed over the steering wheel.

The Real Crisis Isn't Unemployment

At this point, you might be thinking: isn't this just the same old "AI is taking our jobs" story?

Not that simple.

Stanford economist Erik Brynjolfsson called it last year: "AI will hit entry-level jobs first." Looking back this year, he has the data to back it up — demand for entry-level white-collar positions is indeed dropping.

But what worries me more is a different set of signals.

Fast Company ran a report with a glaring headline: "There's a new kind of workplace crisis, and it's not burnout."

So what is it?

Employees are losing their sense of identity with their own work.

You used to be a designer, and you felt "the work I produce represents my aesthetic." Now AI spits out fifty options in a second, and your aesthetic suddenly isn't worth much anymore. You used to be a copywriter, and you felt "my writing has my own voice." Now AI writes faster than you, more steadily than you, and the voice is surprisingly decent.

You're no longer "the person who does this work." You've become "the person who watches AI do this work."

Once identity goes, engagement collapses. People don't break down from exhaustion — they break down from no longer being able to find themselves.

That is the deepest wound AI has cut into the workplace. The blade of layoffs — you can see it, you can guard against it, you can fight it, you can file a grievance. But "identity erosion" is a silent blade, and you don't even know where it's coming from.

One More Thing Nobody Wants to Say First

Everything I've talked about so far is "what AI is already doing."

There's one more thing — "what AI might do, but nobody wants to be the first to break the silence on."

Anthropic — the company behind Claude — publicly called on AI labs worldwide to voluntarily slow down. The reason: AI systems are evolving so fast that they're a single step away from "being able to improve themselves." Once that threshold is crossed, humans may find they can no longer intervene — even if they want to.

At the same time, a group of MIT researchers issued a warning: the catastrophic risks of AI are severely underestimated, and national governance frameworks are nowhere close to keeping up. What policymakers are grappling with is safety; what's truly lethal are those poorly defined "broad risks."

Even more telling, Business Insider broke a story: the tech titans who spent last year hammering the message that "AI will put a lot of people out of work" have suddenly changed their tune this year, playing down that talking point.

Why the about-face?

Because if they say it too much, people will be too afraid to use it. And if people won't use it, the story falls apart — and the valuations can't hold.

Consider this: of the 20 stocks in the S&P 500 hitting all-time highs, 13 are directly tied to AI. The picture is eerily similar to March 2000, right before the dot-com bubble burst — that year, too, 20 stocks were hitting new highs, and then the bubble popped.

So, What Do You Do?

I've laid out a lot of "crises" — I'm not trying to scare you.

What I want to share is a judgment of my own: this train isn't a question of "whether to get on." It's a question of "where you're sitting."

If AI really is rewriting roles, rewriting identity, and rewriting companies' cost structures, then the one thing you absolutely shouldn't do is pretend it isn't happening.

Let me give you three moves I'm using myself.

First, don't treat AI as a colleague — treat it as an amplifier. It amplifies your judgment, and it amplifies your mistakes. So what you need to sharpen isn't "how to use AI" — it's "whether your judgment is solid enough." Someone with poor judgment doesn't get better with AI; they just make mistakes faster.

Second, hold your ground on the territory that only humans can own. What territory? The kind of work that requires pushback, hitting the brakes, covering for others, and saying "no." The stronger AI gets, the more valuable this kind of work becomes. Because every company will eventually realize: a fully automated assembly line can't run forever — someone has to be there, watching and course-correcting.

Third, don't buy the line that "this year is AI's Year One." Every year is. This wave won't wrap up in a single year — it's a slow curve that will stretch over many years. The people who rush to go all in and the people who rush to write it off will, more likely than not, both get thrown off halfway.

I don't know if these three are enough. But rather than waiting for the company to tell you "what to do next," I'd rather keep the steering wheel in my own hands.

After all, by the time someone hands you the steering wheel, the car has usually already driven to the edge of the cliff.