I do three things for a living, and they don't always seem to belong in the same sentence. I'm an intelligence analyst. I run an AI consulting practice. I direct web development for a regional media company. On paper, that's a lot of hats. In practice, they all run through the same set of skills, and lately they've all started running through AI in one way or another.
That's how I ended up on a video call a while back, teaching a small group at Barrister Intelligence how I actually use AI in my investigative work. Not the demo version. The real version. The version where if I get it wrong, somebody's case takes a hit and a real person pays for my mistake.
One of the men in the room had decades of law enforcement experience. He's also a published thriller author, the kind of person who has spent a career on details and getting them right. The kind of person whose skepticism you have to take seriously, because it's been earned.
We were halfway through our second session together when he stopped me to tell me something.
He'd tried AI before. He and his son both had. They hated it. The answers were generic. The writing sounded fake. It got things wrong. He'd written the whole thing off as overhyped.
Then he sat through a class and a half of watching me work.
And he told me, right there in the room, that I had completely changed his mind. Not because I argued with him. Not because I made some pitch about why he was wrong. I didn't even know he'd been skeptical until he told me. He just watched, and what he saw didn't look anything like what he'd tried on his own.
I've been thinking about that moment a lot since.
Because the truth is, I used to want to argue with AI skeptics. I still feel the pull sometimes. I see people in local Facebook groups dismissing AI as garbage, calling everything they don't like "AI slop," telling other people not to bother with it. And there's a part of me that wants to jump in with a counter-example, a side-by-side comparison, a "well actually."
I've stopped doing that. Because what I learned from the gentleman in my class is that the argument doesn't actually work. Even if you win it, you've made an enemy instead of a convert. And more importantly, the skeptics aren't wrong about what they experienced. They tried something, it didn't work, they drew a reasonable conclusion. The fact that the conclusion is incomplete doesn't make it irrational.
Here's what I think is actually happening.
Most people who try AI and walk away disappointed are using it like a vending machine. They walk up, punch in a request, and expect a finished product to drop out the bottom. Write me a blog post about my business. Summarize this document. Make me a logo. When the result comes out generic, bland, or flat-out wrong, they shrug and conclude the machine is broken.
I get the instinct. That's exactly what the hype told them to expect. Push a button, get a miracle.
But a vending machine gives you whatever is already inside it. A real tool does what you guide it to do. A circular saw doesn't cut a straight line on its own. A camera doesn't take a good photo on its own. A scalpel in untrained hands is just a sharp piece of metal that can hurt somebody.
The gentleman in my class watched me do things he wasn't doing. I gave the model context. Who I was, who the work was for, what the output needed to look like, what tone and standards mattered. I gave it real source material, not a vague topic. I pushed back when it gave me weak work and told it specifically why it was weak. I read every word it produced and decided what stayed, what got cut, and what got rewritten in my own voice. I treated it like a sharp junior analyst who needed direction, not a magic oracle.
What he saw, I think, was the thing the hype hides on one side and the cynicism hides on the other. He saw that AI doesn't replace the work. It changes the shape of it. The judgment is still mine. The standards are still mine. The voice is still mine. The accountability is still mine. The AI is in the loop, but it's not driving.
Both crowds, the boosters and the haters, miss this. One thinks the vending machine works and tells everyone to go buy more vending machines. The other tried it, got a stale candy bar, and concluded the whole building is broken. They're reacting to the same misunderstanding from opposite sides.
The building is fine. The vending machine was never the right thing to use.
That's why I don't argue anymore. Arguing puts me in the same frame as the boosters, the ones promising magic. It puts me on the side of the hype, defending it. And that's not where I am. I'm not selling magic. I'm not promising anyone a vending machine that works. I'm doing the opposite, actually. I'm telling people the vending machine doesn't exist and never did, and what's actually here is a tool that takes some skill to operate.
The retired detective in my class didn't need to be argued with. He needed to be shown. And once he saw it, he made up his own mind. That's how grown adults change their minds. Not by being told they're wrong. By watching something true unfold in front of them and drawing their own new conclusion.
So when someone tells me they tried AI and it's all garbage, I don't argue. I just say: I'd love to show you what it looks like when it's not. And then either they take me up on it or they don't, and either way I haven't burned anything down.
I'm still figuring a lot of this out. The work I do crosses three businesses and the through-line is becoming clearer to me every month: I help people use tools they already could have used, but in ways they hadn't figured out yet. Sometimes the tool is WordPress. Sometimes it's an OSINT database. Lately, more and more, it's AI. The principle doesn't change.
Stop expecting the machine to do the work. Start using the tool to do your work faster, sharper, better. Stay in the driver's seat.
That's the whole thing. That's what the man in my class saw. And that's what I keep trying to show everyone else, one ten-minute demo at a time.