The Scarce Assets of Product Management

The artifact was always the visible part

The PM outputs that survive AI drafting are the ones that require standing inside an organization. The PMs who spent a decade perfecting the artifact, the PRD, the roadmap, the deck, bet on the wrong asset, because the artifact was always the part of the job that compresses.

Watch the discussion on Lenny’s, Reforge, First Round. The through-line is that a model can now produce a competent PRD from a prompt, that roadmaps are generated in minutes, that the deck is a solved problem. All true. And it lands as a threat because for a lot of PMs the artifact was the deliverable. It was the thing you were graded on, the thing you showed leadership, the proof you did the work.

That was always a category error. The PRD was the legible residue of a process that happened somewhere else, in rooms and Slack threads and hallway conversations the document never captured. What compresses first is what was always the most compressible: the writing-down. The generation of a clean document from a settled decision was never the scarce part of the job. It only looked scarce because writing a good doc is tedious and most people are bad at it. This echoes the argument that AI made PRD drafting a commodity and left problem framing as the skill that matters.

Being better than average at a tedious, teachable task is not a moat. It’s a wage premium waiting to be arbitraged away.

What the artifact never contained

Open any PRD you’ve ever written and read it as an outsider would. It tells you what to build. It does not tell you why this and not the three alternatives that got killed in review. It doesn’t record what was traded away to get the decision through, or which VP had to be personally walked through the reasoning before the meeting so the meeting didn’t blow up. It doesn’t say that the team tried a version of this in 2023 and it failed for reasons that aren’t written down anywhere, and that half the reason this version is scoped the way it is comes from the scar tissue of that failure.

None of that lives in the document. It can’t. The artifact is the output of a decision. The value is in everything that produced the decision.

A model can generate a plausible roadmap. It cannot tell you that the data team doesn’t trust the attribution numbers, so any roadmap item that leans on those numbers will get quietly deprioritized no matter what the doc says. It cannot tell you that legal will approve the feature only if it ships behind a flag, or that the CEO has a personal aversion to the pricing change you’re about to propose because it burned him at his last company. This is the substance of the job. The document is the part you can hand to a stranger. Everything worth knowing is the part you can’t.

Context is a stock, not a flow

The reason this asset is scarce is that it’s a stock, not a flow.

You don’t acquire organizational context by reading. You accumulate it by being present in one specific organization over time, sitting in the meetings, watching which decisions stuck and which ones unwound, learning who actually holds power versus who holds the title. It compounds the way any stock compounds: slowly, through sustained presence, and it can’t be transferred in a document because it was never fully in a document to begin with.

This is also exactly why it’s a career trap.

A stock that accumulates in one place is illiquid by construction. The PM who has been at the same company for eight years knows where every body is buried, which is enormous leverage inside those walls and worth close to nothing outside them. The context doesn’t travel. Walk into a new org and you’re back to a balance of roughly zero, rebuilding the stock from scratch while the artifact skills, the ones that do travel, are the ones the model just commoditized.

So the durable asset and the portable asset are two different things, and they point in opposite directions.

The uncomfortable trade

Here’s the part that’s hard to sit with. The PM skills you can put on a résumé and carry to any company are, increasingly, the commoditizing ones. Writing tight docs, structuring a roadmap, running a crisp planning process are all teachable, all now assistable by a model, all portable precisely because they don’t depend on any particular organization. Portability and commoditization turn out to be the same property viewed from two angles. A skill that transfers cleanly transfers to everyone.

The skills that survive are the ones that don’t transfer. Knowing how to actually move a decision through your org, whom to convince and in what order, what’s been tried and what it cost, that’s durable because it’s illiquid. It resists compression for the same reason it resists relocation.

This sets up a genuine trade with no clean answer. Optimize for portability and you’re building a skill set that’s worth less every quarter. Optimize for depth and you’re building an asset that’s worth a great deal in exactly one place and near nothing everywhere else.

Depth without renewal is a slow bet on your org staying the same

The obvious move, go deep, stay put, become irreplaceable inside one org, has a failure mode of its own, and it’s worth naming because it’s the strongest objection to everything above.

Context stocks decay. Reorgs happen. The VP who trusted you leaves. The strategy that your accumulated knowledge was built around gets thrown out, and suddenly the eight years of knowing where the bodies were buried is knowledge about a company that no longer exists. The PM who never rotates ends up defending a moat around a castle that keeps getting smaller. Depth without renewal is a slow bet that your one org stays the same shape long enough for the investment to pay off, and organizations rarely oblige.

So the answer isn’t “go deep and never leave,” any more than it’s “stay portable and ride the commodity down.” Both are single-variable answers to a two-variable problem. The PMs who compound are the ones who move deliberately: deep enough to build real context, then out before the stock decays, carrying the one thing that is transferable across orgs. Not the context itself, but the skill of building context fast in a new place.

That meta-skill is the actual scarce asset. It isn’t the artifact, which the model now writes, and it isn’t the local context, which doesn’t travel and eventually rots. It’s the ability to walk into a room where you know nothing and, faster than anyone else, learn who holds power, what failed before, and what everyone is afraid to say out loud. That’s the job the model can’t touch, because the model was never in the room.