I thought I was getting more productive. I was getting more dependent.

When I was writing Human Traits in 2023, I went looking for serious thinking on AI ethics. Not hype. Not fear. Actual philosophical and governance work.

One of the few places I found it was the Montreal AI Ethics Institute. They were asking the hard questions before the hard questions became fashionable: who gets included, who governs, who benefits, and who is left outside the room.

Their recent piece in The AI Ethics Brief asks what Canada's new national AI strategy leaves unsettled. The strategy sets ambitious targets for business adoption. The harder questions: who governs the systems, who owns the value, who can refuse, are largely left for later. The line that stayed with me:

Adoption and benefit are not the same thing.

They mean it at the level of nations and companies. I think it applies to individuals too.

I used AI to help draft early newsletter content. Blog posts. Articles. The process was fast. Suspiciously fast. And then I started reading back what I had produced and something was wrong.

It wasn't mine.

We are being trained to spot the texture of AI-generated writing. Some of it is obvious. Some of it is more nuanced. But the problem I ran into wasn't the words. It was the thinking. The AI had produced output that sounded like content. What it didn't produce was the part that was mine.

I thought I was seeing a benefit. I was taking a shortcut.

Here's the question I eventually had to sit with: why do we form committees? Why were there group projects?

Not to have one person do all the work and distribute it. To produce sharper thinking through friction and divergence of perspective. That's the model that finally unlocked something for me.

AI as a sparring partner. Not a ghostwriter. A challenger. Something that pushes back, surfaces the counter-argument I didn't think of, and asks why I'm framing it that way.

The difference between AI writing for me and AI working with me isn't a subtle one. With me, the piece still needs my thinking. My take. My angle. The mechanics can get a polish. The architecture stays mine.

This conversation — the interview process behind this very issue — is what that shift looks like in practice (em-dashes intentionally included). I wasn't feeding prompts into a text generator. I was being interviewed into a body of work that is mine to approve, edit, and own. That's a different thing entirely.

Praxis: The Authorship Audit

Before you publish, post, or send the next piece of AI-assisted work, stop and answer three questions:

What's mine?
What's my thinking?
What's outsourced?

You don't need to eliminate the outsourced parts. But you need to know where the line is. If you can't locate your own thinking in the piece, it's not yours to claim. If you can, then the AI was doing what a good architect does: handling the load-bearing details on a home you designed.

The goal isn't to use less AI. It's to own more of what you produce.

If you're working on figuring out exactly where that line sits in your own practice, Beyond the Prompt is the cohort I built for it. Next group opens in July. puhala.com/beyond-the-prompt

The piece you're most proud of won't be the one AI made faster. It'll be the one you made better.

– Michael
Founder, The Drop In
& Author of Human Traits — a novel exploring humanity's relationship with AI