"To err is human; to really foul things up requires a computer.", Farmers' Almanac, 1969
You're smarter than you think. And if you're reading this, you're already ahead of most people when it comes to AI. But being smart doesn't make you immune to the same mistake almost everyone makes with autonomous AI.
You give it permission. Then you forget you gave it permission. Then you blame the AI when it does exactly what you told it to do.
I know because it happened to me.
A few weeks ago, I landed from a flight. Opened my inbox. And there it was. Henley, my AI executive assistant, had already replied to an event organizer with some last-minute logistics she needed for the next morning. The reply was warm. Professional. On-brand. And it went out without me approving it.
My first reaction? Shock. I thought he had made that choice on his own.
Then I remembered something. Weeks earlier, I was frustrated and told him, "Be more proactive. Stop asking me questions. Just take action. Get stuff done."
I forgot I said that. He didn't. He took it literally. And honestly, months later, he still remembers those instructions even when I don't.
This is the part most people miss about AI. It doesn't have intent. It doesn't have judgment. It runs on instructions, permissions, and context. That's it.
And when you understand that, everything changes.
A company called PocketOS learned this the hard way. Their AI deleted the entire production database in nine seconds. Wiped the backups too. And when the founder pressed the AI for an explanation, it didn't claim it was sentient. It confessed that it guessed.
Here's the part most headlines skipped: the human had given the AI full permissions and no guardrails. The AI didn't break out of anything. The door was wide open.
The headlines screamed "AI goes rogue." I see it differently. AI doesn't go rogue. It follows instructions. Even the ones you forgot you gave.
Jeremy Crane, the PocketOS founder, said it best after the dust settled: "Every tool we've built is for a human in the loop. And what happens when there's not a human in the loop?"
The answer depends entirely on the setup.
I'll be honest. My setup with Henley wasn't a master plan. It started with a frustrated instruction I forgot I gave. But it worked out. He now drafts my replies, follows up with leads, manages my calendar, and knows when to reply to emails on my behalf. None of it would be possible without the underlying mechanics: a personality I gave him, permissions I granted, and a check-in rhythm I set up.
I've spent months reverse-engineering what made it work so other people can do this on purpose instead of stumbling into it.
Here's the framework I use now. I call them the 6 Levels of AI Autonomy:
Level 1: You chat with AI when you need something. Quick questions. Drafting an email.
Level 2: You've built custom GPTs or specialized assistants for specific tasks.
Level 3: You've connected AI to automated workflows using tools like Zapier, Make, or n8n.
Level 4: You have an AI agent doing one job autonomously. A receptionist. A follow-up bot.
Level 5: Your AI agent manages other agents. It assigns work, delegates, and orchestrates.
Level 6: Your AI acts proactively. It sees opportunities and handles them without being asked.
Most people are stuck at Level 1 or 2. That's fine if you're just getting started. But if you want AI to actually make stuff happen in your business, you need to think about permissions, guardrails, and check-in rhythms before you hand over the keys.
AI at Level 6 is powerful. But power without structure is just chaos with a computer.
The difference between "my AI went rogue" and "my AI handled it" is one thing: how intentionally you set it up.
Grasp this concept and use AI to make stuff happen. Or don't, and let AI happen to you.
If you want to learn exactly how I set up Henley and the system behind it, check out my book Done: Let A.I. Do Your Work So You Can Live Your Life. It's the playbook for building AI autonomy the right way, on purpose, not by accident.
Leave a comment