The CEO’s AI Delusion Isn’t a Bug. It’s a Feature.
Aaron Levie doesn’t hate AI. He’s got 2.7 million followers on X. He writes about headless software like it’s the next coming. He invests in AI startups like it’s his civic duty.
But last week, he called it what it is: psychosis.
Not because the tech is broken. Not because the hype is loud.
Because the people running the companies are losing touch with reality.
They don’t write the code. They don’t review the PRs. They don’t debug the hallucinated API calls that crash the staging environment at 3 a.m. They see a demo. A contract written in 12 seconds. A customer support reply that doesn’t sound like a robot. And they think: ‘This is the future.’
They’re not lying. They’re just… delusional.
And here’s the scary part: they’re not alone.
The entire tech industry is caught in a feedback loop where belief outpaces evidence. We’re not just mistaking AI for productivity. We’re mistaking the idea of productivity for the real thing.
Let me show you what that looks like.
The Layoff Lie: AI Didn’t Fire Them. Finance Did.
In the first five months of 2026, 115,430 people were laid off by tech companies.
That’s nearly as many as all of 2025.
And every single press release? They all said the same thing: ‘We’re optimizing for AI-driven efficiency.’
It’s a lie.
Not because AI isn’t real. But because it’s not the reason.
The numbers don’t lie. Layoffs.fyi shows that the biggest cuts came from companies that had already raised billions, burned through cash, and were now under pressure to show profitability. AI was the perfect scapegoat. Clean. Clinical. Easy to explain to investors.
‘We’re automating 40% of our customer support,’ they said.
But the real story? The CFO had just missed their Q1 margin target. The board was breathing down their necks. So they cut the team that didn’t have a union, didn’t have stock options, and didn’t have a seat at the table.
And then they slapped ‘AI’ on the press release.
This isn’t innovation. It’s accounting.
And it’s working.
Because the market rewards the appearance of efficiency—even when the reality is just cost-cutting.
ClickUp did the same thing.
Zeb Evans laid off 22% of his team. Called it a ‘100x org.’ Said the savings would go back into salary bands.
But here’s what he didn’t say: the people who stayed? They’re now running 3,000 AI agents. Each one of those agents generates output. Each output needs review. Each review takes time.
And guess what?
They’re not getting paid more.
They’re getting more work.
And the worst part?
Gartner’s data shows that 80% of companies using autonomous AI have cut jobs.
But only 12% saw any measurable financial upside.
So why do it?
Because the narrative sells.
And the narrative? It’s not about productivity.
It’s about perception.
The Productivity Paradox: We Feel It. But It’s Not There.
The National Bureau of Economic Research surveyed 750 executives last year.
They asked: ‘Has AI improved your team’s productivity?’
92% said yes.
Then they asked: ‘Has your company’s revenue per employee increased?’
Only 18% could say yes.
Welcome to the productivity paradox.
We feel like we’re getting more done.
But the numbers don’t show it.
Why?
Because AI doesn’t create value. It creates noise.
It generates 100 drafts of a pitch deck. But now you have to choose one.
It writes 50 customer responses. But now you have to approve them.
It drafts the quarterly earnings call. But now you have to fact-check every sentence.
The bottleneck didn’t disappear.
It moved.
From the task to the review.
From the coder to the manager.
From the support rep to the director.
And the people doing the reviewing? They’re exhausted.
A Harvard Business Review study of 200 knowledge workers found that AI tools didn’t reduce their workload.
They intensified it.
People were working longer hours.
More meetings.
More anxiety.
Because now they had to manage not just their own work… but the work of machines.
And here’s the kicker:
The NBER paper says the productivity gains we do see? They’re concentrated in finance and high-skill services.
Not in engineering.
Not in customer support.
Not in product.
In finance.
Where the tasks are structured. Predictable. Linear.
AI doesn’t make your engineers faster.
It makes your CFO’s spreadsheets prettier.
And that’s why the CEO thinks it’s working.
Because the CFO is the one who reports to him.
The Human-AI Team That Doesn’t Work
We’ve all seen the demos.
AI writes a blog post. Human edits it. Together, they produce something amazing.
It’s beautiful.
And it’s a lie.
A meta-analysis of 106 experiments published in Nature Human Behaviour found that human-AI teams perform worse than the better of the two working alone.
Not better.
Worse.
Why?
Because collaboration isn’t magic.
It’s friction.
When a human works with an AI, they don’t just edit. They second-guess. They over-correct. They distrust the machine when it’s right. And trust it when it’s wrong.
A developer gets a code suggestion from Copilot.
It’s wrong.
They fix it.
Then they get another.
It’s right.
They don’t trust it.
They rewrite it.
Now they’ve spent twice as long.
And the AI? It learns from their corrections.
So next time, it gives them the same wrong answer.
It’s not augmentation.
It’s entanglement.
And it’s happening everywhere.
In legal departments.
In marketing.
In design.
The only place it does work? Open-ended creative tasks.
Brainstorming.
Ideation.
The early, messy phase.
But not execution.
Not review.
Not approval.
If you’re using AI to ‘enhance’ your team’s output, you’re probably just slowing them down.
And if you’re measuring success by how many tokens you’ve consumed?
You’re not measuring productivity.
You’re measuring waste.
The Creativity Trap: AI Doesn’t Make You Original. It Makes You Average.
AI generates poetry.
AI writes songs.
AI paints portraits.
It’s dazzling.
But here’s the truth:
AI doesn’t create original ideas.
It remixes them.
A 2025 meta-analysis of 8,214 participants found that when humans used AI to generate ideas, they produced more novel outputs.
But they produced far fewer diverse ones.
Why?
Because AI doesn’t think outside the box.
It thinks inside the most likely box.
It learns from the most popular content.
The most viral tweets.
The most downloaded templates.
The most common marketing slogans.
And then it regurgitates them.
So when your team uses AI to brainstorm a campaign?
You don’t get 100 unique ideas.
You get 100 variations of the same idea.
And that’s dangerous.
Because innovation doesn’t come from the average.
It comes from the outlier.
The weird.
The wrong.
The idea that makes everyone say, ‘That’s not how we do things.’
AI doesn’t help you find that.
It helps you avoid it.
So if you’re using AI to ‘accelerate innovation,’ you’re probably just making your brand more boring.
And if you’re not careful?
You’ll end up with a product that feels like every other product.
Because that’s what AI is optimized for.
The Automation Bias: When You Trust the Machine Too Much
I used to work at a company that automated its incident response.
The AI would detect anomalies.
Alert the team.
And then auto-resolve the most common issues.
It worked great.
Until it didn’t.
One night, the system auto-resolved a critical database timeout.
It thought it was a memory leak.
It wasn’t.
It was a misconfigured firewall.
The AI had never seen that pattern before.
So it guessed.
And it guessed wrong.
The system stayed down for 14 hours.
The team? They didn’t check.
They trusted the AI.
Because it was right 97% of the time.
That’s the automation bias.
When a system is mostly right, we stop questioning it.
We stop looking.
We stop thinking.
And that’s exactly what’s happening in corporate AI adoption.
Executives see AI generate a contract.
It looks perfect.
They sign it.
They don’t read it.
They assume the AI got the fine print.
But the AI didn’t.
It just copied the most common phrasing from 10,000 contracts.
And one of them had a hidden clause.
Now you’re in litigation.
This isn’t science fiction.
It’s happening.
And the worst part?
The more you rely on AI, the worse you get at spotting its mistakes.
Because your brain stops practicing.
You’re not sharpening your judgment.
You’re rusting it.
What CEOs Should Do Instead (And Why They Won’t)
So what’s the fix?
Aaron Levie says it best:
‘Use AI a ton. See what it can and can’t do. Come out the other side with an appreciation for both the upside and the real work.’
Here’s what that looks like:
-
Stop measuring tokens. Start measuring outcomes.
Don’t track how many AI-generated emails your team sends. Track how many customers are still angry after receiving them.
-
Stop automating review. Automate creation.
Let AI draft the first version of the report. But don’t let it approve it.
-
Stop assuming AI makes you faster. Assume it makes you more tired.
Give people mandatory ‘AI-free’ hours. No prompts. No notifications. No assistants.
-
Stop trusting AI for judgment. Use it for volume.
Let it generate 100 variations of a landing page. Then let your team pick the one that feels human.
-
Stop pretending AI is a cost saver. Treat it like a new hire.
A new hire needs training. Mentorship. Feedback. And you’re not giving it any.
But here’s the truth:
None of this will happen.
Because CEOs aren’t trying to fix productivity.
They’re trying to fix perception.
They don’t care if the numbers add up.
They care if the board believes they’re leading the future.
And right now? The future looks like AI.
So they’ll keep pretending.
And the rest of us?
We’ll keep cleaning up the mess.
The Real AI Revolution Isn’t in the Code. It’s in the Culture.
The real revolution isn’t that AI can write code.
It’s that we’ve stopped believing in human judgment.
We’ve outsourced our critical thinking.
To a machine.
That doesn’t understand context.
That doesn’t understand ethics.
That doesn’t understand the cost of a mistake.
And we’re letting it run our companies.
We’re not building a future with AI.
We’re building a future for AI.
And the people who built this? They’re the ones who don’t have to live in it.
They’re the ones who walked away with their bonuses.
While the rest of us are left with burnout.
With anxiety.
With the quiet, creeping fear that we’re not just being replaced.
We’re being erased.
Because if AI can do the work…
…then why do we need to be here?
The answer?
We don’t.
And that’s the real psychosis.
Not believing AI can do the work.
Believing we’re no longer needed to do it.