A peer of mine who's spent decades running large product and engineering organizations was telling me about a conversation he'd had with a product leader. They were talking about whether or not the nature of the product manager role is changing. With AI in the picture, should a product manager still spend their hours writing specs, or move more of that time to the work underneath a good spec: talking to customers, sitting with the data, figuring out what's actually true? The product leader listened to all of it, and then pushed back. "I don't trust anything," they said, "that doesn't come out of my brain."
He didn't argue the point. He kept turning it over for a few days, trying to understand why someone that capable would land there.
What he decided was that the product leader wasn't being precious, or behind. They were protecting craftsmanship. The talent and capability built over decades of doing the work, the kind that's genuinely theirs and can't be faked. The pride was in the human part, and it was earned.
But he also saw the blind spot. The product leader was treating the thinking and the doing as the same act, and they aren't. Even with AI drafting the spec, the judgment behind it stays theirs. The standard it has to clear stays theirs. It still comes out of their brain. It just comes out at a higher order.
It still comes out of their brain. It just comes out at a higher order.
That distinction stuck with me, because most people never find it. They treat AI as all or nothing. Either it does the thinking for them, or they keep it away from anything that matters. The product leader had picked the second. The harder place to stand is the middle, where the tool helps you think more broadly and more completely and the judgment stays yours.
The easy move is to file that product leader under "AI skeptic" and wait for them to come around once the discomfort wears off. That misses what he saw. The product leader's instinct was right. The value a product leader creates does come from their judgment, and guarding that is the correct thing to guard. Their mistake was the next step. They decided that because the value lives in the judgment, their hands have to stay on the artifact, the spec, the doc, the thing they can point to. Those are two different claims, and the second one does not follow from the first.
The other end of the binary makes the opposite mistake. Those product leaders handed the spec to the tool, felt the leverage, and quietly stopped supplying much judgment at all. Speed up, craft down. Neither extreme is the answer. One keeps the judgment and skips the leverage. The other takes the leverage and lets the judgment thin out.
What the skeptic actually needs isn't less skepticism. It's the same judgment, moved up a level. The craftsmanship doesn't vanish when a tool drafts the spec. It relocates to the parts that were always the real work. Which problem is worth solving. What the customer said versus what they meant. Whether the data supports the story or just decorates it. You set the question and the constraints going in, you decide what good looks like and what's out of bounds, and you judge what comes back against a standard the tool never sees. Producing the artifact and exercising the judgment behind it were always two different things. It's just that one person used to do both at once, so the split never showed. Now the tool can produce the artifact, and the judgment is what's left to you. That's the split AI made visible.
This arrangement isn't new, and it didn't arrive with AI. Any product leader who's been at it long enough has done it. A junior PM brings you the spec, the analysis, the first cut. You read it, you push on it and, after a few rounds of edits, you put your name on it. The junior PM did the producing. You, the product leader, supplied the judgment, and you carried the accountability. You got the leverage without giving up the call. That's the shape worth copying.
With one notable difference. When you signed off on that junior PM's work, you weren't the only one who had checked it. They had too, and they cared whether it held up, because their standing with you rode on it. They had skin in the game. AI has none. It faces no consequence for being wrong, and the closest thing to one, you switching to another model, isn't something it feels at all. So its agreement can't carry the same weight, not because the work is poor, but because nothing is on the line for the thing that produced it.
They had skin in the game. AI has none.
Which answers the question that opened the conversation. Does the nature of the role change? It does, but not the way the skeptic feared. The craft doesn't leave. It moves up, to framing and judgment, exactly where he said it already was. There's an inversion buried in that. Handing work to a person usually means owning it a little less. Handing it to AI runs the other way. The tool puts more distance between you and the artifact, and in the same motion makes it more completely yours, because you are now the only human who touched it. That is what higher order actually asks. To own the work more as you make less of it, and to carry, alone, the skepticism that someone else used to share.
Break a Pencil,
PS: Know someone caught between refusing AI and handing it everything? Forward this their way. The hard, useful part is figuring out where your own judgment still belongs.
PPS: There's a harder version of this question I keep circling. If the tool now drafts what the junior PM used to, where do the next senior PMs come from? That's the thread I keep pulling on.