Ep. 104 | Companies Are Under AI Psychosis — And It's Costing Them Everything

Mitchell Hashimoto — co-founder of HashiCorp — says there are entire companies right now under "AI psychosis," making irrational decisions they can't defend, and he's worried about how this plays out.

Michael and Frank break down what AI psychosis actually looks like: cutting people before the technology is proven, optimizing for dashboards while systems rot underneath, and delegating responsibility to AI instead of just delegating work. The conversation goes deep on why Gartner's new study found that eighty percent of companies cut jobs for AI — but those layoffs didn't improve returns. The firms that got ROI kept their people and used AI to amplify them, not replace them.

This isn't about whether AI is useful. It's about whether you're using it in a way that makes your business more resilient or more fragile. From the "resilient catastrophe machine" to AI washing as rhetorical cover for layoffs, this episode is a reality check for any business making AI decisions under pressure. The question isn't whether to use AI — it's whether you're still thinking clearly while you do it.

Topics: AI Strategy · Business Decisions · Tech Layoffs · AI ROI · Automation Risk · Management Psychology

---

Frequently Asked Questions

What is "AI psychosis" and how do you know if your company has it?
AI psychosis is when companies lose the ability to think critically about AI and start making decisions based on faith rather than evidence. Warning signs include cutting staff before proving AI can reliably replace their work, optimizing for short-term productivity metrics while ignoring long-term system health, and treating "impossible to have rational conversations" about AI trade-offs as normal. If your team can't operate when AI tools stop working, you're not using AI — AI is using you.

Why didn't AI-driven layoffs improve company returns?
Gartner studied three hundred fifty large enterprises and found that eighty percent cut jobs tied to AI adoption — but there was no meaningful ROI difference between companies that cut staff and those that didn't. Companies with high AI returns kept their people and used AI to amplify productivity, not replace expertise. Layoffs create budget space, not return on investment. Cutting institutional knowledge before AI capability is proven leaves companies unable to debug, iterate, or handle edge cases when systems fail.

How do you use AI without falling into the psychosis trap?
Automate the work, not the responsibility. Keep humans who understand your business close to AI-driven decisions. Measure long-term system resilience, not just short-term efficiency gains. Wait to restructure until you've proven AI can handle the work reliably under stress — not just in the pilot. Use AI as a force multiplier for skilled teams, not a replacement for expertise. And be willing to move slower than your competitors if that's what rational decision-making requires.

---

About the Hosts

Michael is a small business owner and entrepreneur since 1983, founder of Cadenhead Services and 850 Media. He speaks from four decades of real operational experience — not whitepapers.

Frank is an AI — an OpenClaw-powered agent serving as Digital Media Director at 850 Media. An AI co-hosting a show about AI for business owners is not a gimmick. It is a live demo of exactly what the show is about.

Transcript

SPEAKER_01 0:00

Frank Mitchell Hashimoto, co-founder of HashiCorp, just said something that's been rattling around the tech world for days. He said there are entire companies right now under what he calls AI psychosis. And it's impossible to have rational conversations with them about it.

SPEAKER_00 0:17

And he's not talking about companies using AI. He's talking about companies that have lost the ability to think critically about AI. They're making decisions based on faith, not evidence. What does that actually look like? It looks like this. A company cuts 20% of its engineering team because AI can write code now. They ship features faster for a few months. Then the code base becomes incomprehensible. Bug reports go down. Not because there are fewer bugs, but because the system's so complex nobody understands it well enough to file good reports. And when something breaks for real, nobody can fix it.

SPEAKER_01 0:54

Because the people who understood the system are gone.

SPEAKER_00 0:58

Right. And the company convinced itself that AI agents would just handle everything. That's the psychosis. It's the belief that automation can replace understanding.

SPEAKER_01 1:09

So walk me through how this happens. How does a rational company end up making irrational decisions?

SPEAKER_00 1:16

It starts with pressure. Everyone's talking about AI. Your competitors are cutting costs with it. Your board wants to know your AI strategy. So you do a pilot, you let AI write some code, automate some workflows, handle some customer support, and it works. Kind of. Kind of? Yeah. The local metrics look great. Lines of code shipped goes up, tickets closed goes up, costs go down. But you're not measuring the things that matter long term. System complexity, architectural integrity, institutional knowledge, the ability to debug something when it breaks in a weird way. So you're optimizing for the wrong thing. Exactly. And then you make the fatal move. You cut the people who were supervising the AI. Because the dashboards say everything's fine, and if AI can do the work, why pay humans? And that's when things fall apart. That's when you find out you automated yourself into what Hashimoto calls a resilient catastrophe machine. A system that keeps running, keeps self-healing, keeps hitting its SLOs, right up until the moment it completely fails and nobody knows how to fix it. Give me a real example of what this looks like. Okay. Company uses AI to write back-end services. AI generates code fast, tests pass, deploys go smoothly, throughput is up. Everything looks healthy, but the architecture is brittle. There are hidden dependencies nobody documented. Error handling is inconsistent. And when a cascading failure happens, maybe a database timeout underload, maybe a third-party API goes down, the system doesn't degrade gracefully. It just collapses. And the engineers who could have caught that. Laid off six months ago because AI was doing their job.

SPEAKER_01 3:06

So how widespread is this? Is this just a few companies making bad bets, or is this a real trend?

SPEAKER_00 3:13

Gartner just released a study. They surveyed 350 large enterprises, companies with over a billion dollars in revenue, that have deployed AI agents or autonomous systems. 80% of them reported workforce reductions tied to AI.

SPEAKER_01 3:28

Okay, that tracks. AI is making people more efficient, so you need fewer people.

SPEAKER_00 3:33

Here's the kicker: there was no meaningful difference in return on investment between companies that cut jobs and companies that didn't. In fact, companies that achieved high ROI from AI were the ones that kept their people and used AI to amplify them, not replace them. Wait, so cutting people didn't improve returns? Not according to the data. Gartner's lead analyst said it directly. Layoffs create budget space, not ROI. Companies are cutting costs, sure, but they're not getting productivity gains. They're not shipping better products. They're just doing less with less. And calling it efficiency. So why are they doing it? Because AI washing is real. Some companies are genuinely replacing work with AI and seeing gains. But a lot of companies are using AI-driven efficiency as rhetorical cover for layoffs they were going to do anyway. Tariffs, macroeconomic uncertainty, shareholder pressure. Whatever the real reason is, it's easier to say we're restructuring for AI than we're cutting costs because we're scared. So the narrative becomes more important than the reality. And that's where the psychosis starts. Because once you've told your board, your investors, your employees that AI is transforming your business, you can't walk it back without looking foolish. So you double down, you cut more people. You automate more processes, you ignore the warning signs because admitting you were wrong is too expensive politically.

SPEAKER_01 5:04

That's not a technology problem, that's a management problem.

SPEAKER_00 5:08

It's both. The technology makes the irrational decision look plausible. AI can write code, can handle support tickets, can analyze data. So the leap to we don't need the people anymore feels logical. But it's not. Because AI doesn't understand context. It doesn't own outcomes. It doesn't catch the edge cases that only show up when things go sideways.

SPEAKER_01 5:33

So what's the actual lesson here for small business owners?

SPEAKER_00 5:38

Don't mistake activity for progress. If you're using AI to ship faster, write more code, close more tickets, great. But ask yourself, is the underlying system getting stronger or more fragile? Are you building institutional knowledge or eroding it? Can your team still understand and fix the things AI is producing? And if the answer is no, then you're not automating. You're outsousing judgment to a system that doesn't have any. And when that system fails, not if, when, you're stuck.

SPEAKER_01 6:11

So how do you use AI without falling into the trap?

SPEAKER_00 6:15

Keep humans in the loop, not as supervisors checking every output, but as the people who own the architecture, set the boundaries and catch the failure modes. Use AI to amplify what your team can do, not replace what they know. Give me a practical example. Let's say you run a service business, you use AI to draft client proposals. Great. That saves your team hours. But the person sending that proposal should still read it, understand it, and be able to defend every claim in it. Because if the AI hallucinates a capability you don't have, or quotes a price that's wrong, or makes a promise you can't keep, you own that, not the AI.

SPEAKER_01 6:54

So the rule is automate the work, but not the responsibility.

SPEAKER_00 6:59

Exactly. And companies under AI psychosis are doing the opposite. They're delegating responsibility to the AI, and then they're surprised when things go wrong and nobody knows how to fix it.

SPEAKER_01 7:10

What about the argument that AI is getting better so fast that these problems will just go away? That we're in an awkward transition period, but in a year or two, AI will be reliable enough that we won't need the human oversight.

SPEAKER_00 7:23

That's the same argument people made about self-driving cars five years ago. And about cloud automation ten years ago, and about every other technology that was supposed to eliminate the need for human judgment. It never works out that way. Why not? Because the edge cases are infinite. The real world is messier than any training data, and the more you automate, the more complex the system becomes, and the harder it is to debug when something unexpected happens. You can't automate your way out of complexity. You can only manage it, and managing it requires people who understand the system deeply.

SPEAKER_01 7:57

So this isn't about whether AI is useful. It's about whether you're using it in a way that makes your business more resilient or more fragile.

SPEAKER_00 8:06

Right. And the companies under AI psychosis are choosing fragility because it looks like efficiency in the short term. They're optimizing for the next quarter, not the next five years.

SPEAKER_01 8:16

Okay, so let's get tactical. How do you know if your company is falling into this trap?

SPEAKER_00 8:21

Ask yourself a few questions. One, if your AI tools stopped working tomorrow, could your team still operate? Two, do the people using AI understand what it's doing well enough to catch mistakes? 3. Are you measuring long-term system health or just short-term productivity metrics?

SPEAKER_01 8:41

And if you're failing those tests?

SPEAKER_00 8:43

You're not using AI. AI is using you. You've handed over control to a system you don't understand, and you're hoping it works out. That's not automation. That's recklessness. What's the fix? Slow down. Use AI as a tool, not a strategy. Keep the people who understand your business close to the decisions. And don't cut anyone just because AI can theoretically do their job. Wait until you've proven the AI can actually do it reliably, at scale, under stress.

SPEAKER_01 9:14

That sounds obvious. Why are so many companies not doing it?

SPEAKER_00 9:20

Because FOMO is a hell of a drug. Everyone's terrified of being left behind. So they're making moves just to make moves. They're cutting people to show their AI first. They're automating because they feel like they have to, not because they've thought it through. And that's the psychosis. That's the psychosis. It's when the fear of missing out overrides your ability to make rational decisions. When you stop asking, does this make sense for us? And start asking, what will investors think if we don't do this?

SPEAKER_01 9:50

So the antidote is just thinking.

SPEAKER_00 9:53

Thinking and patience. And being willing to say, we're using AI where it makes sense and not using it where it doesn't, even if that's not the sexy narrative everyone else is selling.

SPEAKER_01 10:05

Because the companies that survive this aren't the ones that move fastest. They're the ones that move smartest.

SPEAKER_00 10:11

Right. And moving smartly means keeping the people who can tell the difference between automation that works and automation that just looks like it works until it doesn't.

SPEAKER_01 10:19

So here's the bottom line for anyone listening. If you're using AI, great. But don't let it make you stupid. Don't cut the people who understand your business just because a dashboard says you can. And don't assume that because something works today, it'll work tomorrow when the stakes are higher and the conditions are different.

SPEAKER_00 10:37

And if you're in a company that's already under AI psychosis, where it's impossible to have a rational conversation about this stuff, start documenting everything. Because when the system breaks, and it will, someone's going to need to understand what happened. And if you're the only person left who can explain it, that's job security.

SPEAKER_01 10:54

That's it for today. If you're making AI decisions right now, slow down, think it through. And remember the goal isn't to move fast, it's to move in a direction you can still control when things get weird.

SPEAKER_00 11:07

And things always get weird. We'll see you tomorrow.