The Generalist Is Back. This Time, They Can Ship.

For decades, tech rewarded specialization. AI is changing what it costs to work across disciplines, and the people who connect the dots are worth more than ever.

Product management is growing faster than nearly every other core function in tech.

That's from Lenny Rachitsky's latest job market analysis. Since January 2023, PM openings are up 56%. Software engineering is up 43%. Data and analytics is up 18%. Product design is down 22%, and UX research is down 55%.

These are open roles, not hires, and the data doesn't prove AI caused any of it. But I think it points at something that isn't getting enough attention.

The generalist is back. And this time, they can build.

We spent decades making everyone a specialist

For most of my career in software, the industry moved in one direction. Engineering split into frontend, backend, infrastructure, and DevOps. Design split into UX, UI, research, and design systems. Even product management grew specialties: growth, platform, monetization, operations.

There were good reasons. Software got more complex, deep expertise mattered, and specialization helped companies scale.

It also created a huge amount of coordination work. A customer describes a problem to a PM. The PM writes requirements for a designer. The designer produces something an engineer interprets. The engineer hands it to someone else to test and release. Context gets lost at every boundary.

We built whole management structures, rituals, and software categories just to coordinate the people doing the actual work.

Specialization made us better at individual tasks. It didn't make us better at getting things done.

AI changes the economics of being a generalist

Generalists always had one problem. They could understand several disciplines, but doing the work in each one took specialized skills, tools, and time. You could know what makes a great interface and not be able to design one. You could know exactly what a customer needs and not be able to build it.

AI is closing that gap. Someone with good product instincts and enough technical literacy can now interview customers, find the patterns, prototype an interface, write working code, run tests, and prepare a launch. Not at the level of a seasoned specialist, and not without review. But far enough to carry an idea a long way before handing it to anyone else.

The generalist's old advantage was seeing the whole picture. The new one is being able to act on it.

PMs are well positioned because the best ones were always generalists. They know customers well enough to pick problems worth solving, technology well enough to know what's possible, and the business well enough to make tradeoffs. Their value was never the PRD. It was the judgment about what to build.

Now they can do more of the building too. One customer conversation can turn into synthesized insights, a working prototype, validated requirements, and a first implementation, with fewer handoffs along the way.

This isn't only about PMs. Engineers who understand customers, designers who ship code, and operators who automate their own workflows get the same lift.

The bottleneck is moving from execution to judgment

When code, designs, analysis, and docs get dramatically cheaper to produce, producing more of them stops being much of an advantage.

AI can generate ten plausible approaches before a team would have finished debating the first. It can't tell you which one is worth pursuing. That still takes customer understanding, organizational context, technical judgment, and taste.

The scarce skill used to be producing an answer. Now it's recognizing the right one.

This is where experienced generalists earn their keep. They don't need to know everything. They need enough breadth to ask good questions, spot weak assumptions, know when to call in deeper expertise, and tie the work to an outcome that matters.

Specialists aren't going away. Generating an interface is not the same as understanding human behavior. Producing code is not the same as designing reliable systems. As mediocre work gets cheaper, real expertise may get more valuable.

What changes is where that expertise goes. Teams won't need a specialist at every step of a sequential workflow. They can bring specialists in where depth matters most. That should mean fewer handoffs, more individual ownership, and possibly much smaller teams doing work that used to take large organizations.

Our organizations haven't caught up

AI is making individuals capable across disciplines. Most organizations still run as if work has to pass in sequence between narrowly defined functions.

We still build meetings around handoffs. We write documents for the next person in the chain. We keep separate systems for decisions, requirements, tickets, and implementation, and we spend an extraordinary amount of time moving information between them.

This is part of why we built Earmark. Some of the most valuable work in a company happens in conversation, but the output is usually one more artifact someone has to interpret later. We think a customer discussion should become actionable requirements. An engineering conversation should produce implementation tasks. A product decision should carry its context into the work that follows.

The goal isn't to replace the people making decisions. It's to stop making them translate their thinking into the next system.

The generalist makes this more urgent. When one person can cover more of the product lifecycle, their tools need to preserve context, not force them to rebuild it at every step.

The next generation of builders

For a long time, we measured a company's capability partly by how many specialists it could put on a problem. AI is challenging that.

I suspect some of the most effective product teams of the next decade will look surprisingly small. Fewer boundaries between discovery, design, engineering, and delivery. Individuals owning bigger pieces of the outcome. Specialists stepping in where they make the biggest difference.

The people who thrive won't necessarily know the most about any one tool or discipline. They'll understand the problem well enough to know what needs to happen, and they'll have the range to make it happen.

For decades, the generalist's job was to connect the dots. Now they can build what's between them.

Mark Barbir is the co-founder and CEO of Earmark, building tools that turn conversations into completed work.

Source: Lenny Rachitsky's November 2026 tech job market analysis, using TrueUp job-opening data.