AI made everyone think they can do everyone else's job

Have we reached the product team standoff? Everyone is armed, everyone is pointing at somebody else, and nobody wants to lower the weapon first.

PMs think they can design and code. Designers think they can ship and set strategy. Engineers think they can own product and the pixels. AI handed each of them a tool that makes another person's job look trivial from the outside, and each of them believed it.

I beg to differ, and not because I am sentimental about job titles.

Tasks are moving. Jobs are not.

The clearest framing I have heard on this came from Marc Andreessen on Lenny's Podcast, where he argued that the atomic unit of work is the task, not the job, and that a job is really a bundle of tasks. His line: "The job persists longer than the individual tasks."

That is the whole thing. Everybody is watching tasks disappear and reporting it as a role disappearing. Those are different events, and confusing them leads teams to make organisational decisions they regret about six months later.

Product team roles are not vanishing in the next five to ten years. They are being re-bundled. The job descriptions will change. Output velocity will go through the roof. The humans behind the work are still very much needed, just for different parts of it.

What each role is actually losing

It helps to be specific, because the panic is always vague and the reality never is.

The product manager is losing artifact production. The spec, the deck, the competitive matrix, the first draft of the release note. What is not going anywhere: deciding what is worth building, and being the person accountable when that call was wrong.

The designer is losing production work. Twelve variants of a screen used to be a week. It is now an afternoon. What is not going anywhere: knowing which of the twelve actually serves the person using it, and why the other eleven do not.

The engineer is losing boilerplate. Scaffolding, glue code, the tedious middle. What is not going anywhere: judging what a thing will cost to run, to maintain, and to change in eighteen months when the person who generated it has left.

Look at that list and the pattern is hard to miss. What moved out of each role is the part that could be described precisely enough to hand off. What stayed is the judgement.

Why the standoff is expensive

Here is the practical damage. A product decision carries four risks at once: is it valuable, is it usable, is it feasible, is it viable for the business. A product trio exists because those four questions are best answered by three people who see the same decision from different places, at the same time.

The trio was never three skill sets in a room. It is three vantage points on one decision.

So when one person concludes that AI now lets them cover all four, the honest description of what has happened is that you have quietly rebuilt the single-genius model, and you have lost the disagreement that made the decision good. The tooling got better. The blind spot did not move.

And blind spots are not cheap. Roughly 80% of features are rarely or never used. That number was produced by teams who had designers and engineers and product managers all present. It does not improve because one of them can now generate the other two's output.

What actually needs to change

Not the org chart. Three other things.

Write job descriptions around decisions, not deliverables. If a role is defined by the artifacts it produces, AI has already made most of that description obsolete. If it is defined by the calls that person owns and is accountable for, it survives the next five years intact.

Make ownership explicit where it now overlaps. When a PM can produce a design and a designer can ship a change, "whose call is this" stops being obvious. It has to be said out loud once, or it gets litigated every sprint.

Raise the bar on the part that did not get automated. Judgement under uncertainty is now the scarce resource on a product team, and it is the one thing nobody has figured out how to generate. Teams that are still being measured on throughput will keep hiring for the wrong half of the job. That distinction is the same one that separates a delivery team from an empowered product team, and AI has made it much sharper rather than less relevant.

So which role is changing the most?

My honest answer is the product manager, and it is not good news dressed up as bad.

For years, a large part of the PM's day was artifact production. Writing the thing, formatting the thing, distributing the thing. That was never the actual job, but it filled the calendar and it looked like work, and plenty of people built a career on being fast at it.

That half has collapsed. What remains is the harder half: choosing what to build, defending the noes, and carrying the outcome. Which is uncomfortable, because it removes the last place to hide, and it makes very obvious who was doing product management and who was producing documents. It also explains why the mini-CEO framing was always wrong. Influence without authority was the skill before AI. It is closer to the entire job now.

The roles are not collapsing into each other. The responsibilities inside them are being redrawn, and most teams have not done that redrawing deliberately. That is precisely the work of a product model transformation: deciding who owns which call, before the tooling decides it for you by accident.