AI Is in My Way Now

Let me set this up before anyone gets the wrong idea.

I run an IT, cybersecurity, and AI company, and a marketing agency. I'm writing a book about AI for small business. I use these tools every single day, and I tell other owners to use them. I am the last guy who should be complaining about AI.

And yet. AI is in my way now.

Not the technology. The way people are using it. I've watched it happen inside my own walls, and the research says it's happening almost everywhere. It moves through four stages. Almost nobody stops it at stage one, because stage one feels too small to complain about. That's the whole problem. Stay with me.

Stage one: the messages

Somebody gets excited about AI. Two weeks later, every message they send has been run through it. You can tell. Everybody can tell.

The messages get longer. They pick up a polish the person never had. They open by summarizing what you just said and close with an offer to help further.

People who used to answer in one sentence now answer in four paragraphs, and three of them say nothing. You raise a real concern and what comes back isn't their answer. It's a machine's answer with their name on it.

Do the math. Ten seconds to generate. Three minutes to read. Across a team, the time one person saved just became a cost on everybody else.

And when I ask you what you think, I want to know what you think. If I wanted a chatbot's opinion, I'd ask the chatbot. It's faster and it doesn't pretend to work for me.

Stanford and BetterUp Labs actually named this last year: workslop. AI-generated work that looks finished but moves nothing forward. Forty percent of workers say they got some in the last month. And 42 percent of the people receiving it said they trusted the sender less afterward.

Read that again. You think you're leveling up. Your coworkers are quietly writing you off.

Stage two: the trust

Nobody announces that they've stopped trusting a colleague. They just start reading that person's messages faster. Double-checking their numbers. Routing around them on anything that matters.

And the person sending AI-written everything never gets told. Because who wants that conversation?

So they keep doing it, and the gap keeps getting wider. It started at stage one. Nobody said anything.

Stage three: the systems

Here's where it stops being annoying and starts costing money.

The same enthusiasm that had people running their emails through AI has them building apps with it. Good. Build things. I mean that.

But then the apps start talking to real systems. Someone hooks up an API. Someone sets up a sync into the database the whole company runs on. Nobody reviews it, because it never went through anybody.

And now company names are misaligned in the CRM. Records don't match. Data other people depend on is quietly wrong, because of a tool somebody built on a Tuesday afternoon with an AI coding assistant.

I've spent 30 years telling businesses not to let this happen with their servers. Don't let one person be the only one who knows how it works. Don't push to production without a second set of eyes.

Every one of those rules applies here, and almost nobody is applying them. Why? Because it doesn't feel like IT. It feels like a guy being productive.

Stage four: the dependency

Eventually somebody tells you they can't do part of their job because their app is broken.

Stop and think about that. We did that job for years before anyone wrote an app for it. The job never needed the tool. The tool was supposed to serve the job.

But that's not what got built. The person put themselves, and their homemade tool, between the company and something the company already knew how to do.

Tool breaks, work stops. Nobody falls back to the old way, because in their head the old way is gone.

Padded messages. Lost trust. Corrupted data. And then one day AI is the reason revenue didn't happen. Every stage is bigger than the last, and it all traces back to some long emails nobody wanted to mention.

We are not anti-AI

I need to say this plainly, because somebody's going to read this and decide I'm yelling at the future.

I'm not. I'm writing a book telling small businesses to use AI more, not less. The best people I know use it constantly, and I usually can't tell.

Because they use it to sharpen their own thinking, and then they send me their own words.

The Stanford researchers have a phrase for this: pilots and passengers. Pilots steer the tool and add their own judgment. Passengers let it carry them and skip the hard part.

Coworkers can tell the difference every single time. My problem was never the tool. My problem is passengers.

And here's the part I wish more people would sit with. The pilot gets more valuable every month, because the tool is multiplying their judgment. The passenger gets less valuable every month, and they don't feel it happening.

Think about what you're actually paid for. It's not typing. It's knowing the client, knowing which of the four paragraphs matters, knowing when the tool is wrong. That's the part nobody can subscribe to.

And when everything you send has been run through AI, that part disappears from view. Not because you lost it. Because you hid it.

If I can't find you in your work, I've learned nothing about you and I've trusted you a little less. The tool doesn't take that hit. You do.

Use AI to make your judgment more visible, not to replace it. That's the whole game.

This is what “rogue” actually means

I'm calling the book Rogue AI, and people hear that and picture some sci-fi scenario. No.

Rogue AI is what you just read. It's AI that got into your business through the side door, one excited employee at a time, with no plan and nobody deciding how much of it belongs in each job.

And that last part is what owners skip. The right amount of AI in a sales role is not the right amount in bookkeeping, and neither one is the right amount for the person touching your client data.

Somebody has to make those calls on purpose. Set the guardrails. Check that they're holding. That's an AI initiative. It's a real project with a real owner, not a Google Chat message that says “go play with ChatGPT.”

It's the work we're doing at Directive and BriteAgency right now, in our own operations and our own marketing. And it's the work most businesses haven't done.

So here are my rules

Take them.

• Use AI to think. Use yourself to talk. Draft with it. Argue with it. Then close it and write the message yourself.
• You own everything you send. There is no “the AI wrote that.”
• Nothing touches shared systems without a second set of eyes. Build in a sandbox. The second it writes to data other people depend on, somebody else reviews it first.
• Every automation needs a fallback. If your tool breaks and the work stops, you didn't build a tool. You built a single point of failure with a chatbot in front of it.
• Judge the output, not the tool. I don't care if you used AI. I care whether what you gave me was right, clear, and yours.

Set the rules while it still feels too small to matter

Looking back, I should have said something the first week the emails got long. It felt like nagging. It was the moment.

So here's what I'm asking you to do. If people on your team suddenly sound like a press release, that's stage one. The next three are already on their way.

Have the conversation now, while it's still about writing style and not about your database.

AI is the best tool I've picked up in 30 years of running a business. I just want it out of my way.