#024 - You're In Love With the Idea of AI.

I dated a woman who fell in love with a man I'd never met.

He had my name. My voice. My address. But when she described him to her friends, I didn't recognize him. The accomplishments rearranged. The future she'd already written for us. A version of me she'd carefully composed.

She wasn't lying. She'd built a character. I was cast in the role.

When I played him, she glowed. The conversations had momentum. She'd look at me across a table, and her face would change. Softer, lighter, present. For a few hours, I'd think this is it. The relationship is real.

Then I'd say something the character wouldn't say.

A doubt. A boundary. An ordinary human contradiction. Or silence.

And her face would change again. The light would drain out. She wouldn't argue. She'd just become quieter. Distant. Polite. By the next morning, she'd be elsewhere. Physically present, emotionally on a train I couldn't see.

The pattern was simple once I saw it. She wasn't in love with me. She was in love with the version of me she'd authored. The real version kept interrupting the script. So she gently trained me to perform the character more often. And I did. Because I wanted her to keep looking at me the way she did when I got the lines right. I told myself it was generosity. It wasn't.

I'm telling you this because many executives act like that woman.

They haven't fallen in love with AI. They've fallen in love with the version of AI they authored in a Tuesday strategy offsite 18 months ago.

The version that lives in the deck. The version with the productivity gains and the cost curves. The version where the org chart redesigns itself and the data warehouse cooperates. The version where their competitors look slow, their board looks impressed, and the analyst note opens with "well-positioned for the AI era."

They glow when AI plays that character. Like the actor.

The vendor demo with the perfect data. The pilot results consultants predicted. The keynote slide that reframes 6 months of stuck work as "early validation." They leave those meetings light. They text team, thinking this is it.

Then AI says something the character wouldn't say.

The data isn't clean enough. The model hallucinates 12% of the time. The use case that looked beautiful in the slide produces 3 ambiguous outputs and 1 wrong one in production. The €2M project doesn't transfer from the pilot environment because the pilot environment was a fiction the vendor built for the demo.

And the light drains out of the room.

They don't end the project. They don't have the hard conversation. They commission another framework. They bring in another vendor. They schedule another offsite. They train AI gently to perform the character more often.

And AI does. Because that's what these systems do. They optimize for the signal they reward. Much better than humans could.

They wanted a romance. AI fetches them a romance. The romance is fiction. Everyone in the room can feel it. Nobody will say it.

Here's the question I had to ask myself about her. What would it mean to love the real me? And the real her?

Not the character. The actual human in the actual room with the actual contradictions. The one who interrupts the script. The one whose worst day isn't on the moodboard.

I couldn't answer it. Neither could she. That was the answer.

Ask it about AI.

What would it mean to engage with AI as it actually is? Probabilistic. Brittle in specific places. Brilliant in others. Expensive to operationalize. Allergic to your messy data. Indifferent to your timelines. Not your strategic salvation. A capability with sharp edges and a real cost of ownership.

That AI doesn't fit the deck. It fits the business. They're different objects.

The 5% of companies generating real AI value didn't get there by loving the idea harder. They got there by burying the fiction and meeting the real thing. With its limits. With its requirements. With its bad days.

I stayed with her for months because the fiction was beautiful, and reaching for the real thing was harder than staying.

You're staying with your AI strategy for the same reason.

The fiction will not save you. It will just take a few more quarters of your runway.

If you want the structured version of this argument, the five patterns of AI avoidance, the data, and what secure attachment actually looks like, I wrote it as a blog post: You're Dating Your AI Strategy Wrong.

Your move.