
My first job was at Nokia. The web server sat under my desk. I learned Photoshop so I could edit the images of the phones, and I built an image-map navigation in JavaScript for the main Nokia website, a world first at the time. I wrote it, shipped it, ran it, and heard from the people who used it. I did every job, until the volume of work grew too large and agencies had to come in.
That job has a name now. PostHog has written a whole Product Engineer Handbook for what it calls a new reality: Product engineers talk to users. They decide what to build. They own pricing, revenue, and user experience. They support customers directly. Job ads for AI product engineers are appearing in Barcelona this month, and the title is spreading fast.
I recognise the job. I just don’t recognise it as new. The product engineer is regression to the mean: for most of the history of software, the people who built products also ran them. The long chain of specialists between an idea and a shipped change is the exception, and AI is removing what made it necessary.
What Agile assumed
In 2001, seventeen software people wrote the Agile Manifesto. Its principles read like a description of my desk at Nokia: Business people and developers must work together daily throughout the project. Build projects around motivated individuals. Give them the environment and support they need, and trust them to get the job done. The best architectures, requirements, and designs emerge from self-organizing teams.
None of that is easy with large groups of narrow specialists. A team that writes its own requirements, designs its own architecture and works with the business every day needs people who can do more than one job. The manifesto assumed them, which tells you its authors were them. They were writing against the first wave of specialisation, the heavy, document-driven processes that had already split the work into stages, and defending what small teams of builders who talked to the people they were building for had always done well.
Work grew to fill the people
Expertise got more expensive, but the bigger change was that work got laborious, very often unnecessarily so.
In 1955, the historian C. Northcote Parkinson wrote in The Economist that work expands so as to fill the time available for its completion. His essay was really about headcount: officials multiply subordinates, and the work grows to keep them all busy, whatever the real workload. Software proved him right. As the industry grew, the sequence went like this:
- building became cumbersome, with more tools, more process and more steps between an idea and a release;
- each new step got its own role, and each role its own specialist;
- the handoffs between specialists became work in their own right, with their own meetings, documents and coordinators;
- everyone stayed extremely busy, while speed and impact went down.
Nobody designed it like that: it is what happens when every problem is solved by adding a role. A designer who draws, a PM who writes, an engineer who builds, a tester who checks, and nobody who owns the result.
Startups kept the old way, until the money arrived
Startups were always the counter-example. A small group building fast can’t afford a chain of handoffs, so everyone does several jobs and the product moves. Then the funding arrives, and most of them lose it. They hire to a plan rather than to a need.
A Barcelona scale-up I know set itself a target to double its headcount every year, with no clear idea of what the new people would do. It grew fast, then spent the following years cutting more than half of its peak headcount. The problem was the plan: it treated growth in headcount as progress, and the work expanded to fill the people.
Some companies chose differently. Lenny Rachitsky’s account of how Linear builds product is short: No product managers, just a head of product. PM duties are distributed across engineering and design. Linear stayed small on purpose, and its engineers do the job PostHog is now writing a handbook about.
What AI removes
The laborious parts are exactly what AI is good at. An engineer can now sketch an interface, a designer can prototype working code, and a product manager can read the codebase and write a precise change. Deep expertise still matters, but the cost of contributing outside your own speciality has collapsed. The barrier that made multi-skilled people rare is falling, and the mean is coming back.
That is why I don’t share the excitement about a new role. The product engineer is the person the industry spent years splitting into parts, being put back together. I have written about the career side of this, the T-shaped professional and the Renaissance professional. The organisational side is harder, because it means removing roles, and it connects directly with the question of which product manager AI is replacing.
Specialists still matter
Specialisation did happen for good reasons. Systems got bigger, and security, scale and data really are hard. Nobody can be an expert in everything. A platform engineer, a security specialist or a data engineer is not going away, and the product engineer works with them, not instead of them. What goes is the chain: the sequence of handoffs where each person does one slice and nobody owns the outcome. There is a fair worry, too, that product engineer is a way to get three jobs for one salary. If a job ad really means that, it is a pay problem, and it should be called one.
Before you post a product engineer job ad, count the handoffs between an idea and a shipped change in your team today. Then ask which of them exist because of a gap in skills and which exist because of the org chart. The first kind you can hire for. The second kind you can simply remove, and the people on either side of each handoff you remove will start doing the job I did at Nokia.
Nobody invented the product engineer. The industry just stopped forbidding them.
Leave a Reply