Who Creates Software Is Changing. Not Everyone Will Make the Shift.

Three conversations last month. Same shift, three different responses.

A product manager at an enterprise company told us she'd started working alongside AI agents to draft product specs. Not as a novelty. As her actual workflow. She was producing in hours what used to take weeks. Her reaction: "I'm never going back."

A senior developer at the same company heard about it and shrugged. "That's fine for specs," he said. "But the real work is still code." He was using a copilot to write functions faster. Same moment in history. Same access to the same capabilities. Completely different understanding of what had changed.

A third person, a domain expert outside the company, had been building a prototype on her own with AI agents. She'd hit a wall, but she wasn't discouraged. She was energized. "I built more in a weekend than I thought possible," she said. "Now I need to figure out what comes next."

Three people. Same tools. Already diverging.

The pattern nobody wants to talk about

Every major shift in computing has produced winners and losers. Not because the technology was exclusive. It was available to everyone. The dividing line was never access. It was always the same thing: who changed how they worked, and who didn't.

Here's the uncomfortable part. The people best positioned to lead each shift were usually the last to make it.

Kodak invented the digital camera. Not a competitor. Not a startup. Kodak's own engineer, Steve Sasson, built one in a lab in 1975 and showed it to management. Their response: "That's cute, but don't tell anyone." They had the technology, the distribution, the brand, the customer relationships. They had everything except the willingness to become something other than a film company. By the time they tried, it was too late.

The Swiss watch industry invented quartz movement technology, then dismissed it. Mechanical watchmaking wasn't just their product. It was their craft, their identity, their pride. Seiko took the technology the Swiss had invented and nearly destroyed them with it.

Blockbuster had more stores, more customers, and more capital than Netflix. They even had an early online offering. But they couldn't let go of the store, the late fees, the model that made them Blockbuster. Netflix didn't just change delivery. It changed what it meant to be in the business.

The pattern repeats. The people who mastered the old model had the most to lose by embracing the new one. Not just financially (though sometimes that too). Psychologically. Their expertise, their status, their sense of who they were... all built on a way of working that was about to become optional.

The shift happening right now is following the same pattern. And the resistance isn't coming from people who lack capability. It's coming from people whose identity is built on the old model.

Three roles, three versions of the same trap

The developer who measures value in code. For decades, writing code was the bottleneck. If you could write it well, you were valuable. Write it fast and well? Exceptional. That was real. It was earned. And now the skill that made you exceptional is being commoditized. Not eliminated. Commoditized. The developer who responds by prompting faster is optimizing the old model. The one who becomes an architect of human-agent systems, who brings judgment and product thinking to a team of collaborating agents... that person is making the shift. The most important skill was always judgment. But the identity built around code output is hard to release.

The organization that confuses handoff with collaboration. Most companies building software follow the same process: someone writes requirements, someone else builds to spec, someone reviews the output. They call this collaboration. It isn't. It's a relay race where the baton carries less context at every exchange. The shift asks organizations to put the domain expert in the build, not adjacent to it. That means restructuring teams, rethinking roles, changing who's in the room. Most organizations will optimize the relay race instead. Add AI to the existing workflow. Call it transformation. Get the same results, slightly faster.

The domain expert who's been told they "can't build." This one is different. The trap here isn't clinging to an old identity. It's believing a limiting one. For years, domain experts were told that building software required technical skills they didn't have. That was true then. It's less true every month. The expert who's already prototyping, already creating with AI agents, already pushing past what they thought possible... they're closer to the shift than almost anyone. The only thing stopping some of them is the belief that they need someone's permission to be in the work. They don't.

What the shift actually asks

Here's what's easy to miss. The shift isn't asking anyone to become something they're not. It's asking them to become more of what they already are.

The developer's deepest value was never typing speed. It was understanding systems, seeing the architecture, knowing what to build and why. Those skills matter more now, not less. But only if the developer stops defining their value by lines of code.

The domain expert's deepest value was always their knowledge of the problem space. Now that knowledge can directly shape the solution. They're not becoming developers. They're becoming co-creators, participating in the build in ways that weren't possible two years ago.

The organization's deepest value was always its ability to coordinate talent toward a goal. The shift doesn't eliminate that. It transforms the coordination from sequential handoff to genuine collaboration, with humans and agents working together across every stage.

No one's role shrinks. Every role transforms. But transformation requires letting go of the old shape first. That's where people get stuck.

The window

This is what makes the pattern most uncomfortable. It's not just that some people adapt faster. It's that the window for adapting closes.

In every previous shift, the early movers didn't just get a head start. They shaped the landscape. The companies that figured out the web didn't simply arrive first. They defined what "good" looked like for everyone who came after. By the time the rest showed up, the rules had been written without them.

The same dynamic is playing out now. The people and organizations figuring out how to work with agents, not just prompt them faster, are developing instincts, patterns, and ways of collaborating that will define this era. They're building the muscle memory of a new way of working while others are still debating whether the old way needs updating.

Who creates software is changing. The tools made it possible. The willingness to change made it real.

Not everyone will make the shift. Not because they can't. Because they won't.