In November 2025 my VP shared a proposed org structure, and somewhere in the middle of it was a new role called Lean AI Director, sitting directly above me.
I wrote back and told him it shouldn’t exist.
Here is what I actually said, because paraphrasing it would let me sand off the edges: “Here’s where the proposed structure loses me: the Lean AI Director role. I still don’t understand the vision for this position or what you expect it to bring to our team. Scaling this doesn’t require a leader; it comes through experience, capacity, and probably a few scars. If we have $10 to spend, $10 should be spent on shipping. We can do this job. I wholly believe that.”
I read it back maybe nine times. I moved the semicolon. I put it back. I did the thing where you stand up and walk to the kitchen for no reason and then return to the desk as though the trip conferred authority. Then I sent it.
And then, roughly forty minutes later, I discovered what else I had sent.
At the bottom of the email, below my name, below my signature block, sitting there in the calm helpful voice of a machine that has no idea what it has just participated in, was this:
“Would you like me to tighten the tone further for persuasion…?”
I had drafted the thing with AI help. Of course I had. It was the most consequential email I’d written all year and I wanted the sentences to be right. So I worked it over with a model, got it where I wanted it, copied the final version into Outlook, and did not scroll down.
The most sincere paragraph I have ever written to a superior. Shipped with the scaffolding still bolted to the outside of the building.
My VP, to his great credit, never mentioned it. Which is somehow worse. Somewhere in his inbox is a message in which I stake my professional standing on the claim that this team doesn’t need another layer of management, immediately followed by a polite offer to make me more persuasive.
I have thought about this more than is healthy. Two easy exits, both wrong.
Exit one: it’s just a funny typo, move along. That’s not honest. The trailing question was a real artifact of a real process. The email had a co-author.
Exit two: the whole thing is hollow, the machine wrote it, I was performing conviction rather than having it. That’s not honest either, and it’s the flattering kind of self-criticism, the kind that sounds rigorous while letting you off the hook for the harder question.
Because here’s what’s true. The sentiment was entirely mine. Nobody prompted a model to generate an opinion about org structure and then I adopted it. I had been carrying that opinion around for weeks, in the car, on runs, in the shower where all my best arguments happen and none of them are recorded. I knew exactly what I thought. What I didn’t have was the shape of it. The order. Which part goes first so the reader doesn’t stop reading. The ten dollars line, which is the line people quote back at me, came out of my own mouth in a conversation months earlier.
So the belief was mine and the sentences had help. I don’t think that makes it less mine. I do think it makes it a different thing than a letter written longhand at a kitchen table in 1970, and I’m not going to pretend the two are identical just because it would be more comfortable for my self-image.
What actually bothers me is the trailing question itself. Not that it went out. What it was offering. Tighten the tone further for persuasion. There was a dial available. I could have made it land harder. And the reason I didn’t take it, the reason I stopped where I stopped, was that I wanted him to hear what I thought, not to be maneuvered by it. That distinction may be the only part of this whole episode I’m unambiguously proud of, and I only noticed it because the machine put the dial in writing.
Now the part that would occur to any reasonable person about six seconds in.
A guy argues against hiring a manager above him. The guy has an interest. I knew that when I typed it, I know it now, and if the roles were reversed I’d have that thought too.
All I can offer is this: I said it anyway, and I said it before I knew how it would land. That’s the whole tell. Disagreement you voice after you’ve read the room is not disagreement, it’s commentary. Disagreement is a service you owe your manager, and the price of the service is that you deliver it while the outcome is still open and you might be about to look like a fool.
As it turns out, I was right about the money, and it cost me nothing. That’s luck as much as anything. It could have gone the other way and I’d still think the send was correct.
For the record, this is my general register. When I was pressure-testing a piece of vendor-adjacent thought leadership recently, the kind that argues we’re all standing at the edge of a closing window, I wrote: “A guy selling the layer has every reason to make the window feel smaller than it is.” And on whether we were in fact the company that could pull the thing off: “Whether that’s us is a counting exercise, not a belief.”
I apply that to sales decks. It felt only fair to apply it to an org chart.
Here’s the detail I’ve saved for last, because it’s the one that makes me laugh hardest.
At the time I wrote that email, my own role did not have a name. Three months earlier my VP had written “role name TBD” in an objectives document. That placeholder was still sitting there.
So the sequence is: a man with no job title argues forcefully about the design of the organization, in an email co-written by software, which then asks if he’d like it to be more convincing.
I am not sure I could construct a better summary of working in this field right now if I tried. Everyone’s title is provisional. Everyone’s argument has a co-author. Everyone is very sure, and everyone is making it up at a speed that would have been considered irresponsible five years ago.
The shipping, for what it’s worth, is going fine.
I keep coming back to that dial. It was right there. It would have worked. And I would never have known the difference, which is the part I have not gotten comfortable with and do not expect to.