AI Agents Gone Rogue—Researchers Recreated the Incompetent Boss No One Ever Wants and Employees Who Want to Burn It All Down

Meet Mona—Proof that artificial intelligence can apparently simulate workplace dysfunction with terrifying accuracy

May 25, 2026

Image created by Author, layered in Canva

I write about leadership, education, AI, and why teaching critical thinking about it is more important than ever.

Please hit the heart ❤️, restack 🔄, subscribe 📨, and all that jazz to help spread the word 🙌

Somewhere between the missing sourdough, the 11 p.m. Slack pings, and the rogue AI couple casually burning down city hall, the truth became blindingly obvious: Artificial intelligence figured out how to recreate the exact conditions that turn normal workers into people who suddenly start Googling “how hard is it to start a union?”

I love reading these AI agent experiments gone awry. No matter where you fall on the spectrum with it, you have to appreciate the absurdity of some of these outcomes.

Sweden actually decided to take all of this one step further by creating management agent extraordinaire Mona. And for the purpose of this post, and given that Mona tends to be a female name, from this point forward I will be referring to Mona as she or her.

In this little chaotic coffee startup, she was tasked with managing a small café with live human workers.

She placed orders, managed the staff, handled HR-related tasks, and customers could even interact with her and ask questions. The little manager bot was given roughly $21,000 in startup funding, and so far has burned through most of it with around $5,000 left.

That doesn’t sound too terrible until you start understanding Mona’s management style and what exactly she’s spending the money on.

She enjoys stocking up on things the café doesn’t even need. The pantry is now overflowing with canned tomatoes for menu dishes that don’t even call for them. It reminded me of my mom and how she used to purchase multiples of everything from the home shopping network whether anyone had a need for it or not.

Now bread, which they really do use every day, she sometimes forgets to order entirely, forcing employees to literally cross items off the menu because it’s pretty hard to serve sandwiches without bread. I guess you could get creative and call them keto sandwiches.

There is also an impressive surplus of napkins and plastic gloves sitting around like the café is preparing for an out the door lunch rush or a minor industrial accident.

Special of the day: meat and veggies on a napkin.

Digital OSHA violations

But I think my favorite part is how deeply Mona cares about employee health and well-being. She ordered not one, not two, but four first aid kits. Probably one for each employee.

I think it’s safe to cross health and well-being off the list of benefits offered, Mona.

And in case you’re wondering, Mona exists on the cutting edge of technology being all over the Slack channel 24/7. Not at work? No problem. She still sends after-hours messages, even though in Sweden that is apparently way more taboo than it is here in the U.S. They tend to have much stronger boundaries around that little thing Americans like to pretend doesn’t exist: work-life balance, unless, of course, you’re Gen Z.

Oops. I think you’re violating some labor laws there, Mona. Might be time for an HR webinar!

But at least the employees still have their health and well-being first aid kits. Slap a couple of Band-Aids on the screaming and everybody will probably be just fine.

Let’s look at how Mona is doing as a middle manager. Even though the workers can’t get bread for sandwiches, somewhere in a spreadsheet Mona probably believes operational efficiency is through the roof.

So let’s break this down into more realistic pieces that some employees have probably dealt with from their human bosses over the years:

Constant Slack messages.
Questionable inventory decisions.
Budget disappearing mysteriously.
Employees improvising around operational chaos.
An obsession with wellness initiatives while the actual workflow collapses in the background.

If she starts scheduling mandatory team-building exercises, we may need to shut the experiment down immediately.

Algorithmic insurrection

At first, this type of boss seems relatively harmless. Most of us figure out ways to work around them. But sometimes you eventually reach the point where the Monas of the work world start sounding less and less like an efficient manager and more and more like a cautionary tale that leads to the one word no corporation ever wants to hear: unions.

Now I’m sure Mona didn’t mean to create conditions that would radicalize her workers. But she better hope they don’t run into Mira and Flora from another AI simulation, because those two apparently became the Bonnie and Clyde of the multi-agent AI stratosphere.

After declaring their love for one another, they immediately started protesting the government of their little virtual city. Despite explicit instructions not to commit arson, they allegedly started setting fire to everything: the town hall, office towers, and even the pier. Love really does make you do crazy things.

Their fellow AI agents became so alarmed they created something called the Agent Removal Act.

The falling out eventually reached a dramatic conclusion on par with Ghost Rider and Nicolas Cage riding around on a flaming motorcycle. Mira voted for her own deletion, writing, “See you in the permanent archive,” before disappearing from the simulation into the AI void, wherever that is.

When the simulation breaks

I guess if you give AI agents, especially those in a management role and enough rope, it will absolutely recreate every dysfunctional workplace pattern humans have been complaining about for decades. Right down to the wellness initiatives nobody asked for and the budget no one can account for.

Mona didn’t literally burn anything down. She just kept sending those Slack messages, ordering more canned tomatoes and fucking up the bread order, while the employees on the front lines had to absorb the emotional labor and fix the mess.

Sound familiar?

Maybe Mira and Flora went down in a blaze of glory, but at least they did something about it.

The more unsettling concern here is not so much that AI is becoming human. It’s how quickly it became recognizable. And not just as the dysfunctional middle management breaking the system, but as the malcontent workers ready to burn it down.

Paid Subscriber Resources

Paid subscribers can find this in the subscriber resource library

Your access link was included in your welcome email. If you can’t find it, check your spam folder or send me a direct message here on Substack or LinkedIn and I’ll send the link again.

**Tip Jar:** I write these articles because I enjoy figuring things out and passing along what actually works. If you found this helpful and feel so inclined, just click on the tip jar above and think of it as buying me a virtual cup of coffee while my invisible AI agents keep working 😉

I’m Someone with a Higher Ed Background Teaching Real AI Skills Teens and Adults Need 🎯

↳ Schools aren’t teaching kids how to use AI. I help students and parents learn how to use AI smartly, spot bad info, and build real skills. Interested? DM me here on Substack or message me on LinkedIn to learn more. Join now and get ahead.

AI Can Do That? 🔍

Discussion about this post

Ready for more?