The Approval Rule: Why Nothing Leaves Without My OK

I use AI assistants every day. They help me sort email, draft posts, and get work ready for me to look at.

But there's one rule I don't bend: nothing leaves without my OK. No email, no post, no message. If it goes out with my name on it, I said yes first.

Here's why, and how I check the work before I trust it.

Fast Is Only Good if It's Right

AI tools move fast. That's the point. But fast in the wrong direction is still the wrong direction.

A wrong reply to a customer, a post with a bad number in it, or a message sent to the wrong person can take a lot longer to clean up than it would have taken to check. A quick review costs me a few minutes. Skipping it could cost me a relationship.

Drafts First, Always

Every assistant I use makes drafts. Nothing comes out as a final on the first pass.

That changes how I think about the work. I'm not asking for a tool to do my job. I'm asking it to get the job ready for me. Then I decide.

What I Check Before I OK Something

Here's what I look at before I say yes:

• Is it true? Numbers, dates, prices, and names get checked against the source. If I can't find where a number came from, it doesn't go out.
• Does the math add up? Totals, percentages, and ranges get a second look.
• Does it sound like me? If I wouldn't say it out loud to someone, I change it.
• Is it going to the right place? Right person, right channel, right time.
• Does the timeline hold up? Who did what, and when. If a claim says something happened, I make sure it could have.
• Would I be OK if this was wrong? If the answer is no, I slow down and check again.

AI isn't 100% accurate. It can be confidently wrong. That's not a knock on it. It's just how it works, and it's why the human piece still matters.

The Day the Rule Earned Its Keep

Here's a real one from October 26.

One of my bots put together a topic sheet for my show. It was a solid draft. Clear, organized, ready to talk from. And one line in it said that my dev bot wrote the prompts for version 1.1 of my notes app.

It sounded right. It fit the story. It just wasn't true.

Version 1.1 was built before that bot existed. It couldn't have written anything for it. The bot that made the sheet had connected two true things, a dev bot and a notes app, into one false one.

I caught it because I read the sheet before I used it. If I hadn't, I'd have said it out loud on a live show, to real people, as a fact about my own work.

That's the kind of mistake that worries me most. Not the obvious typo. The smooth, believable line that fits so well nobody questions it. It's also the kind of thing that would have been easy to correct before, and awkward to correct after.

Why Smooth Mistakes Slip Through

A bot doesn't know when it's guessing. It writes a guess in the same calm voice it uses for a fact.

That means the mistakes don't look like mistakes. They look like everything else on the page. The only reliable way I've found to catch them is to check the claims that matter against what I actually know, or against the source.

Timelines are a big one. Who did what, and when. A bot can see two related things and assume one caused the other, even when the dates don't line up. That's why "does the timeline hold up?" is on my list above.

A Person Approves Anything Risky

Some things get extra care no matter what:

• Anything that sends or posts in my name
• Money
• Deleting things
• Anything that's hard to undo

For all of those, a person signs off. Right now that person is me.

Setting Up Your Own Approval Rule

You don't need special software for this. You need a few decisions, made ahead of time.

1. Decide what counts as "going out." Emails, posts, messages, anything a customer or the public will see. Write the list down.
2. Make drafts the default. Tell every tool you use that its job is to get something ready, not to send it.
3. Pick your checks. Start with the list above, then add the ones that fit your work. If you deal with prices, prices get checked. If you deal with dates, dates get checked.
4. Name the person who says yes. In a small business, that's probably you. Say it plainly, so nobody, human or bot, assumes otherwise.
5. Keep a short note of what you catch. When the same kind of mistake shows up twice, add it to your checks.

It feels like extra work for about a week. After that, it's just how you work.

It's a Habit, Not a Bottleneck

You might think this slows everything down. It mostly doesn't.

The assistant does the legwork. I review a finished draft instead of starting from a blank page. That's usually faster than doing it all myself, and I still get the final say.

Technology is a powerful tool. But it's agnostic. It isn't good or bad on its own. The person using it still decides whether it does the job. The approval rule is how I make sure that person is me.

The takeaway: Let AI do the legwork, but keep the yes. Drafts first, check the facts and the math, and a person approves anything that sends, spends, deletes, or can't be undone.

Try It This Week

Pick one thing an AI tool does for you right now. Before you use what it gives you, run it through the six checks above. See what you catch.

If you're just starting, What to Automate First in a Small Business, and What to Never Automate covers which jobs to keep for yourself.

Here's my question: what's one thing you'd never let go out without checking it yourself?