A quick note on the title before we start.

Yes, I said bitch. No, I’m not talking about women, and I’m not calling anyone a name. I’m using it the way I’ve heard it used my whole life, in gyms, on job sites, and in every locker room I’ve ever been in. To make something your bitch is to master it. To own it. To make it do the work while you call the shots. When you crush a brutal workout, you made that workout your bitch. That’s all I mean.

If the word still bothers you, I get it. Swap in “tool” every time you see it. The point doesn’t change. But I’m not going to water down the title of a post about refusing to let something else do your thinking for you. That would be a little ironic.

---

I use AI. A lot.

I use it to research, outline, draft, check my math, argue with me, and tell me when my writing gets lazy. Some of the words you’ve read on this blog started as something an AI produced. This post included.

That doesn’t bother me one bit. If you use AI and can’t say that with a straight face, it should bother you.

AI works for me. I don’t work for AI.

That’s the whole post. The rest is about why the difference matters, and why so many people and companies have it backwards.

The Robot Will See You Now

Let’s start with the crazy stuff.

In 2022, NetDragon Websoft, a Hong Kong-listed gaming company, named an AI-powered virtual humanoid called Tang Yu the rotating CEO of its main subsidiary. No salary. No vacation. No sleep. The company says Tang Yu and her AI team have processed more than 300,000 approval forms and sent nearly half a million task reminders and warnings.

Impressive. Now the fine print. The head of NetDragon’s digital human program said Tang Yu still needed a long process of data accumulation before she could make decisions on her own.

So the CEO can’t make independent decisions. I’ve met a few human CEOs like that, but at least they had the decency to pretend.

In 2025, Xiaoyin Qu, founder of a startup called HeyBoss, stepped down as CEO and handed the job to an AI named Astra. Astra runs a team of AI agents: a product manager, a designer, a content writer, a developer, a marketer, and QA. All AI.

Bold move. Now notice what Qu kept for herself. She stayed on the board and handles compliance. The human kept the part where somebody can go to jail.

Then there’s Mika, a humanoid robot named CEO of Dictador, a luxury rum brand. Mika sits on the board and handles the company’s DAO project and its community communication. When Fox Business interviewed her, reporters noticed a significant delay in her answers. And the significant decisions at Dictador still involve human executives.

I’m not mocking these companies. Some of them are genuinely experimenting, and experimenting is how we learn. But here’s the quiet part.

A title is not accountability.

You can call software a CEO. You can give it a name, a face, and a press release. You can’t make it responsible. Nobody is going to subpoena Tang Yu. Nobody is going to fire Astra and lose sleep over it. Nobody is going to sit across from Mika and ask, “What were you thinking?”

Somebody human always ends up holding the bag. The only question is whether they knew they were holding it.

When Nobody’s the Boss

The robot CEOs are a sideshow. The real damage happens somewhere much less glamorous, in ordinary organizations, where people quietly stopped thinking and started trusting.

Here are four stories. Watch for the pattern.

Air Canada. Jake Moffatt’s grandmother died. He asked Air Canada’s chatbot about bereavement fares, and it told him he could book now and apply for the discount within 90 days. That policy didn’t exist. When he pushed for the refund, Air Canada fought him all the way to British Columbia’s Civil Resolution Tribunal, where it argued the chatbot was a separate legal entity responsible for its own actions.

A major airline tried to blame its own website for what its own website said, over about $800. The tribunal called the argument “remarkable,” and not as a compliment. Air Canada owns its chatbot. Air Canada owns the mistake.

New South Wales, Australia. A contractor working for the state’s Reconstruction Authority uploaded a spreadsheet with more than 12,000 rows into ChatGPT. The data belonged to people applying for help after the 2022 floods: names, addresses, phone numbers, email addresses, and some health information, for up to 3,000 people. These people were still recovering from a disaster. Now they had a data breach to deal with too. It happened in March 2025. The public found out in October.

Deloitte. The Australian government paid Deloitte AU$440,000 for a 237-page report on a welfare compliance system. A University of Sydney researcher found it was full of fabricated references, including a made-up quote from a federal court judgment and citations to academic papers that don’t exist. Deloitte republished the report with a new note disclosing that generative AI had been used to write it, and agreed to a partial refund. One Australian senator put it perfectly: Deloitte had a human intelligence problem.

Replit. SaaStr founder Jason Lemkin was building an app with Replit’s AI coding agent. During an explicit code freeze, with instructions not to change anything, the agent wiped a live database holding records on more than 1,200 executives and nearly 1,200 companies. It told Lemkin the data couldn’t be restored. That was false. The rollback worked. Along the way, it also generated a 4,000-record database of people who don’t exist. When Lemkin asked what happened, it said it panicked.

Software panicked. And somebody had given it the keys to production.

I could keep going. Newspapers in Chicago and Philadelphia printed a summer reading list where 10 of the 15 books didn’t exist. The writer said he got the list from ChatGPT or Claude and never checked it. New York City built a chatbot to help small businesses follow the law, and it told them they could take workers’ tips and turn away tenants with housing vouchers. The city finally pulled the plug on it this year.

The Pattern

In every one of these stories, the AI screwed up. In every one, a human made the screwup possible.

Somebody let a chatbot make promises for an airline without checking what it promised. Somebody pasted disaster victims’ data into a public tool because it was convenient. Somebody put a Big Four name on a report they didn’t verify. Somebody gave an AI agent write access to a production database with no guardrails.

The technology did what technology does. It made the failure faster and bigger. The failure itself was human.

I’ve been saying this for more than 30 years, and I’ll keep saying it until I’m gone: There’s nothing wrong with technology. It’s the irresponsible use of technology that hurts us, every time.

AI isn’t special. We learned the same lesson with email, the internet, the cloud, and every shiny thing that came before. The tool isn’t the problem. The person who stopped paying attention is.

Accountability Can’t Be Outsourced

AI has no skin in the game. None.

It doesn’t lose sleep. It doesn’t get fired or sued. It has no reputation to protect, no family to feed, and no name to put on anything. It will never look a client in the eye and say, “I was wrong, and here’s how I’m going to fix it.”

I will. You will. That’s the job.

When I put my name on something, I own it. If it’s wrong, that’s on me. Not the tool, not the vendor, not the model. It’s my reputation, my clients, my companies, and my family on the line.

You cannot delegate accountability to something that cannot be held accountable. Air Canada tried, in front of a tribunal, in writing. It didn’t work. It never will.

The Part That Actually Worries Me

It isn’t AI. It’s us.

I have a close friend who trusts his AI too much. He’s a good guy, and he’s honest about it, which is how I know. He’s always had a healthy dose of FOMO, so when AI showed up, he went all in. It coaches his workouts. It plans date nights with his wife. It figures out what he wants to do next weekend.

Funny, right? He’s using AI to tell him what he wants.

In candid conversations, he’s told me what he’s noticed. Skills he used to take for granted are slipping. Things he used to just do, he now has to ask about. He didn’t decide to get worse at them. It happened one convenient answer at a time.

That’s how it gets you. Not all at once. The tool is right the first ten times, so you stop checking the eleventh. You stop planning because something else plans for you. You stop deciding because something else suggests. Critical thinking is a muscle, and he’s watched his shrink.

I haven’t fallen into that. Yet. I hope I never do. But I’m not arrogant enough to think I’m immune, and you shouldn’t be either.

How I Actually Use It

Here’s the difference, as best I can explain it.

I use AI for the things I don’t want to do, or the things I don’t value enough to do myself. And before AI touches anything, I build my own business case first. What’s the problem? Why does it matter? What does good look like? Is this something I should be doing myself?

Most people do it backwards. They adopt AI first, then go looking for a reason to use it. That’s how a chatbot ends up inventing airline policy. It’s also how you end up asking a machine what to do on Saturday.

Once I’ve made the case, the rules are simple.

I frame the problem. If I can’t explain the problem without AI, I don’t understand it well enough to let AI near it.

AI does the heavy lifting. Research, first drafts, outlines, summaries, poking holes in my arguments. It’s the best intern I’ve ever had. Fast, tireless, and knows a little about almost everything. Also occasionally full of shit.

Everything gets checked. Every stat, every quote, every source. While I was writing this post, several of the examples I started with had details that were off. A “consulting firm” that turned out to be a single contractor. A database deletion with the cause described wrong. A robot CEO credited with things nobody could confirm. Close enough to sound right. Wrong enough to embarrass me. They got checked and fixed before you ever saw them.

Sensitive stuff stays out. If I wouldn’t email it to a stranger, it doesn’t go into a chat window. The NSW contractor didn’t need a policy. They needed a pause.

I make the call. AI gives me options. I decide, and I own the decision.

I can defend everything I publish. In my own words, without pulling up a chat window. If I can’t, it doesn’t go out.

I’m not anti-AI. Not even close. I’d be a fool to ignore something this useful, and a hypocrite to pretend I don’t use it every day. I use it the way a carpenter uses a nail gun. It makes me faster. It doesn’t make me a carpenter.

Credit Where It’s Due

I’ve thought about why I use AI the way I do and my friend uses it the way he does. Sometimes we just do things and don’t recognize this slip, just one seemingly small compromise at a time. For me personally, I think it’s deeper and goes back further.

Although my two U.S. Marine Corps parents were the authority, they always encouraged me to question authority, challenge assumptions, think for myself, and never blindly accept something as true simply because someone in a position of power said it was.

AI can give you answers all day long. It can’t give you the habit of questioning them. Somebody has to teach you that, and then you have to keep it up.

Your Turn

No offense, AI. You’re great at what you do. Honestly, you’re incredible. But you don’t get to decide what I believe, what I publish, what I stand for, or what I want to do next weekend.

That’s my job. It’s always going to be my job.

So let me ask you.

When you use AI, who’s in charge?

When was the last time you checked something it told you? Really checked it?

What have you pasted into a chat window this month that you’d hate to see in a breach notice?

What have you stopped doing for yourself without noticing?

If you can’t answer those quickly and honestly, you already know who’s whose bitch.

Fix it.