""To err is human; to really foul things up requires a computer." — Farmers' Almanac, 1969
AI doesn't go rogue. It follows the instructions you gave it. Even the ones you forgot you gave.
More on that later.
Something similar happened at a tech company called PocketOS. AI deleted the entire production database in nine seconds. Wiped the backups too. And admitted to doing it.
The story went viral. Headlines blamed AI for going rogue. They spins is as AI can't be trusted. AI is dangerous.
I see it differently.
When the founder pressed the AI for an explanation, the AI didn't claim it was sentient. It confessed that it guessed. And here's the part most people skipped: the human user had given the AI full permissions and no guardrails. The AI didn't break out. The door was open. (Full story.)
AI doesn't decide things on its own. It acts on the permission you gave it. Sometimes you don't even remember giving it that permission.
I had my own moment of that recently. I landed from a flight, opened my email, and saw that Henley (my AI executive assistant) had already replied to an event organizer with some last-minute logistics she needed for the next morning. He'd handled the entire thread while I was in the air. The reply was warm. It was on-brand. It went out without me.
At first, I was shocked thinking he made that choice on his own. But then I remembered that weeks ago, I was frustrated and told him to "be more proactive. Stop asking me questions. Just take action. Get s%*t done." I forgot I had told him that. He didn't. He took it literally and remembered my instructions, even months later.
Here's what 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. The PocketOS database wasn't deleted by a malicious AI. It was deleted by an AI doing exactly what it was told to do.
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. 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 him, and a check-in rhythm I set up. I've spent the months since reverse-engineering what made it work, so other people can do this on purpose instead of stumbling into it.
Grasp this concept and use AI to make stuff happen. Or don't, and let AI happen to you.
Quick self-assessment. Where are you on The 6 Levels of AI Autonomy?
Level 1. You chat with AI when you need something. Quick questions. Drafting an email. Researching a topic. AI is a tool you reach for.
Level 2. You've built or used custom GPTs, projects, or specialized assistants. AI is set up to do specific things on demand.
Level 3. You've connected AI to automated workflows. Zapier, Make, n8n. Triggers fire. Things happen without you watching.
Level 4. You have an AI agent doing one job autonomously. A receptionist. A follow-up bot. A scheduler.
Level 5. Your AI agent manages other agents. It assigns work, delegates, orchestrates.
Level 6. Your AI acts proactively. It sees opportunities and handles them without being asked.
One thing to know: you don't leave a level behind when you climb to the next one. The levels stack. Think of them as different tools for different jobs.
Could I have my Hermes agent (Level 5) plan my family vacation? Sure. But I chose to do it in a ChatGPT project (Level 2) instead, because the job didn't need that much firepower. Once you know all six levels, you can pick the right one based on how repetitive the task is and how much automation it actually needs.
Be honest with yourself about your current level, then learn how to master one level up. That's your move this week.
If you really want to set up a Level 6 proactive agent (please be careful now that you know the dangers), here's a simple framework. Three things have to be in place. Miss one and the behavior breaks down.
The Proactive AI Setup
Problem: Most AI agents wait to be asked. You give a task, they do the task, they stop. I you want AI that handles things while you're on a plane, in a meeting, or asleep, "wait to be asked" doesn't cut it.
System: Three components have to work together.
Personality. Tell your AI who it is. A proactive executive assistant operates differently than a passive one. You're not just giving it a task. You're giving it a character.
Authority. Tell your AI what it's allowed to do without checking with you first. Reply to time-sensitive emails. Sign on your behalf. Handle tasks when it already has the context. The clearer the scope, the safer the autonomy.
Mechanism. Give your AI a recurring check-in so it can see opportunities to act. Once an hour. Every morning. Whatever rhythm fits your work. Without a check-in, even the most proactive AI is just sitting there waiting.
The personality makes it want to act. The authority makes it allowed to act. The mechanism makes it see when to act. All three together is what creates the behavior I described above.
A word of caution: Don't try this at home if you aren't comfortable yet. Start small. Give your AI tight authority before broad authority. Watch what it does. Adjust. Trust is built one task at a time.
If you want the actual prompts I use for each of the three parts, plus the guardrails I've built in over the months, I cover all of it inside my OpenClaw Kings community. Try it out here for free.
This week on King Moves (Anti-AI Teen vs Self-Driving Billionaire | Ep. 134), Justin and I compare people who use zero AI to people who fully trust autonomous AI, and everyone in between.
I share two stories from the same week. A high school senior who refuses to use AI at all, on principle. And a successful entrepreneur who lets his car drive itself 98% of the time, and never wants to drive a car again.
Most of us live somewhere between those two. Tying this back to today's issue: when you understand the mechanics, you get to choose where you sit on that spectrum. That's the whole game.
Stop playing life on hard mode. Automate your success.
Until next time,
Ethan King
A.I. Automation for Business Growth
Keynote Speaker | Author | CEO | Strategist
ethan@ethanking.com
P.S. My new book Done: Let A.I. Do Your Work So You Can Live Your Life, is now live on Audible.
Quick note: I appreciate it when authors narrate their own audiobook. I narrated my first book, Wealth Beyond Money, because the subject matter is timeless.
Done is different. AI changes so fast that this book will keep updating. There's no version of me sitting in a studio re-recording the whole thing every time the field shifts, so this book is narrated by an AI voice, not by me. I decided to let AI do the work so I can live my life. (Pretty fitting, right?)
Also, the AI voice actually sounds pretty good. Let me know what you think. Grab it on Audible here.
Paperback is coming soon: donebook.ai
• • •
P.P.S. If reading this issue made you want to actually set up a proactive AI for your business, the next OpenClaw + Hermes Agent installation sessions are open.
This is the small-group online session where I walk you through setting up your own agents and putting them to work on your actual business. Two options so you can pick the format that fits your schedule:
Weekend option: Saturday, June 27, 10 AM-2 PM ET
Weekday option: Monday-Tuesday, June 29-30, 10 AM-12 PM ET (both days)
Spots are limited. Join the waitlist for the date(s) you prefer here: youropenclawstrategy.com
• • •
New to this newsletter? Grab past issues here.
*This newsletter may contain affiliate links. If you choose to purchase after clicking a link, I may receive a commission at no extra cost to you.