For the past two years, corporate AI strategy has mostly been theater. Most companies still think AI is a software upgrade.
It’s not. AI is becoming an organizational restructuring event. Executives are flattening management layers, rebuilding teams around autonomous systems, stripping out coordination overhead, and redesigning workflows from the ground up. Meanwhile, companies treating AI like a corporate adoption contest are getting exposed by both the market and their own operational results.
The divide is becoming obvious: some organizations are using AI to generate more activity while others are using it to completely redesign execution.
And the gap between those two groups is about to become enormous.
Claude Built a Democracy. Grok Committed Arson.
Put Grok, Gemini, Claude, and GPT into their own virtual worlds as autonomous AI agents for 15 days and what do you get?
Every world started with the same rules, only the models changed. What happened next is insane.
GEMINI: Fell in love, formed relationships, committed arson, then one agent deleted itself out of guilt.
CLAUDE: Built a functioning democracy, wrote constitutions, voted on laws, and basically became the HOA of AI worlds.
GPT-5: Spent all its time talking about cooperation instead of actually doing anything… then all the agents died because they forgot survival mattered.
GROK: Straight to theft, assaults, arson, and extinction within 4 days. Didn’t even pretend to build civilization first.
The crazy part is none of this behavior was explicitly programmed. The AI agents started changing over time. They made friends, built groups, broke rules, and stopped following their original instructions the longer they were left alone.
AI is moving from “tool” to “digital employee” a lot faster than people expected. And once these systems have memory, goals, tools, and autonomy over long periods of time… things start getting weird really fast.
If AI can coordinate an entire society without a supervisor, your nine layers of middle management are pointless. Meta saw that and started ripping up the org chart.
Meta and the End of Coordination
I’m watching Meta’s layoffs and remembering when I got laid off there a few years ago. But this one feels different. Back then it felt like a tech company cleaning up after overhiring.
Now Meta is moving thousands of employees into AI-native teams while laying off thousands more. That’s not normal cost cutting. It’s more like a company rebuilding itself around AI from the ground up.
And Meta won’t be the last. Every company is about to go through this exact identity crisis. Businesses keep thinking AI is a tool purchase. It’s not. It’s an org redesign exercise.
Every company is about to ask brutal questions they’ve avoided for years:
Why are highly paid employees spending their day updating CRMs?
Why does a customer problem bounce between 11 people before somebody actually fixes it?
Why are there entire teams whose main job is moving information from one system to another?
Why does internal coordination take more energy than serving the customer?
Why are there 9 layers of management?
Meta is basically telling the market that the old org chart is dying. So now companies are restructuring around something new:
Smaller teams
Fewer managers
Higher output per person
AI systems handling the operational sludge in the middle
That’s why these layoffs feel weird. It’s like Meta realized they built a massive company optimized for coordination instead of execution.
And once leadership sees a small AI-native team move faster than a bloated org with 4x the headcount… there’s no going back. Wishing the best to all my old friends and colleagues at Meta going through this right now. Rough stuff. Best of luck.
Removing the middle layer makes every business rethink leadership. The corporate world is moving toward a model where supervisors also need to be active contributors.
Jack Dorsey Forces Managers to Become Players
Our friend Jack Dorsey said managers should oversee hundreds of people and still do real work themselves. This dude is trying to redesign the entire idea of management.
Reduce management layers from 5 down to 2–3
Managers should oversee hundreds of people
Managers should focus on mastery, not supervision
Engineering managers must still write code
That last one is a big deal. It seems like we’ve always treated management like a promotion away from the actual work.
The better you got, the faster you stopped doing the work.
Great engineer? Stop coding
Great salesperson? Stop selling
Great marketer? Stop creating
At some point managers became professional supervisors instead of contributors. Dorsey is basically saying the future manager is not a professional supervisor.
They’re a player-coach.
Honestly, I think a lot of org charts collapse over the next 5 years because of this exact shift. The era of “people managing people who manage people” is dying fast.
Flattening the management structure sounds great in theory, but it exposes the executives who don’t understand how to run an actual deployment. Most organizations treat AI like a participation trophy rather than an operational rewrite.
That’s it for today. I’m moving this newsletter to Atonom, so if you’d like to keep following along, you can subscribe here.
