Where AI-Displaced Engineers Actually Go: Data, History, and What to Do Next

In Part 1, I explored why professional identity threat (not technical skill) predicts who struggles with AI tools. The developers most at risk aren't the least capable. They're the ones whose sense of self is tightly coupled to practices being automated.

But that raises practical questions: What does history actually show about professional transitions? Where do displaced developers go? And what capabilities complement AI rather than compete with it?

The answers are more nuanced than either doom or dismissal.

What History Actually Shows

Every time a technology has transformed a profession, predictions of wholesale elimination have proven wrong. But something does happen, and understanding the pattern matters more than either optimism or doom.

When CAD transformed architecture beginning in the 1980s, hand drafting (which had required 15-20 extremely skilled specialists) consolidated into generalist positions requiring both design skills and technical proficiency. One veteran described going back to school at age 40+ for retraining. Another noted doing the work of at least five people: designer, typesetter, scanner operator, camera operator, film stripper.

The pattern wasn't elimination. It was bifurcation. Younger designers who embraced the change often pushed their firms to adopt faster. Older professionals who resisted found their specific skills marginalized. But the profession itself didn't vanish. It transformed into something that required different capabilities. The Bureau of Labor Statistics now projects little employment change for drafters through 2034, with roughly 16,200 annual openings primarily from retirements.

The same story played out in finance after the quantitative revolution. Before 2007, investment banks typically hired MBAs as entry-level analysts. After the financial crisis, firms shifted toward quantitative analysis skills. Quant firms like Jane Street now pay average salaries of $900,000, triple what traditional traders at Goldman earn. The profession bifurcated into those who could work alongside algorithmic systems and those who couldn't.

The cross-cutting finding: technology transforms professions by elevating some roles and marginalizing others. The roles that get elevated share a common characteristic. They sit at the interface between automated capability and irreducibly human judgment.

Where Displaced Developers Actually Go

The 365 Data Science study of 1,157 laid-off tech workers between November 2022 and January 2023 provides hard data on outcomes. About 40% had found new jobs by March 2023. Of those with new positions, 19% joined smaller software development firms, 13% went to internet companies, and most continued careers in tech, often at similar or higher levels. According to Karat's data, 61% of displaced developers stayed at the same level when targeting new roles, 27% actually leveled up by moving to smaller organizations, and only 12% interviewed at lower rungs than before.

The pattern is more nuanced than either "everyone's fine" or "the apocalypse is here." Layoffs from 2022-2024 totaled over 400,000 tech workers across thousands of companies. But most stayed in tech. Many found positions quickly. The ones who struggled most, according to the 365 Data Science research, were those in the 40-50 and 50-60 age ranges.

Big Tech headcounts have largely recovered. Meta has 19% more engineers compared to January 2022. Google's engineering headcount increased 16%, Apple's 13%. The hiring has shifted, though. AI engineering roles have exploded since mid-2023. The Pragmatic Engineer newsletter reports that software engineers are finding it relatively straightforward to transition into AI engineering by building applications on top of LLMs.

The off-ramps for those who don't want to adapt exist too. Legacy systems running COBOL handle trillions of transactions across finance, healthcare, and government. Someone has to maintain them, and that expertise is becoming rarer. Regulated industries where AI adoption moves slowly offer refuges for traditional development approaches. Adjacent roles like DevRel, technical writing, and product management absorb people with technical backgrounds who want different relationships with code.

The Capabilities That Complement AI

MIT Sloan's 2025 EPOCH framework analyzed 19,000 work tasks across 950 occupations to identify which human capabilities correlate with both protection from automation and employment growth. They found five clusters, and none of them are what most engineers think of as their core competencies.

• Empathy: emotional intelligence and human connection.

• Presence: physical networking and relationship-building.

• Opinion: ethical judgment and accountability.

• Creativity: imagination, humor, and improvisation.

• Hope: leadership, grit, and initiative despite uncertainty.

All five showed positive association with employment growth from 2016-2024. Hope showed the largest impact.

If you're reading this as an engineer, that list might feel uncomfortable. These aren't skills that show up on technical assessments. They're not what most engineering cultures have historically valued or selected for. But they're precisely what complements AI capability rather than competing with it.

As Hersh Tapadia, CEO at Allstacks, puts it: "The best engineering organizations are sitting there saying, we are leveraging outcomes-first thinking and then building what we're building. The funny thing is that was true pre-AI too. The best engineering organizations were already doing that. They just had a couple extra reps to go through between leveraging that outcome-first thought and getting the code out the door. Now they can close that loop."

The distinction he draws between "rubber ducking mentality" and "agentic mentality" captures something important. The rubber ducking developer uses AI as a sounding board while maintaining full control over code production. The agentic developer treats AI as a team member to be managed, providing context, reviewing output, iterating on results. One protects the old identity. The other builds a new one.

The Uncomfortable Bottom Line

The World Economic Forum's 2025 Future of Jobs Report projects 170 million jobs created and 92 million displaced globally by 2030, a net positive, but cold comfort if you're in the 92 million. The developers who remain will likely be fundamentally different from today's median: more product-minded, more architecture-focused, deeper in domain expertise, more comfortable with ambiguity and supervision.

And there's another factor worth acknowledging. The 20+ year professionals who have developed comfortable routines are often the leaders and decision-makers driving AI adoption in their organizations. That creates a tension: the people setting the pace of transformation may themselves be the ones most invested in existing ways of working. Whether that creates resistance or creates space for thoughtful transition is an open question, and probably varies organization by organization.

Some engineers will adapt. Not because they have better technical skills, but because they've built identities that can evolve. They tolerate the dissonance of working with tools they don't fully trust. They focus on outcomes over methods. They see AI as expanding what they can accomplish rather than diminishing who they are.

Others won't adapt. Not because they lack capability, but because their professional identity is too tightly coupled to practices that are being automated. They'll find niches in legacy systems, transition to adjacent roles, or exit the industry, following the same patterns we've seen in architecture after CAD, finance after quants, law after e-discovery.

If you're reading this and feeling defensive, that's useful information. Not a verdict, but a signal. The defensiveness itself suggests how tightly bound your identity is to current practices. Which means you have a choice: protect that identity and find a niche where it's still valued, or begin the uncomfortable work of building a new one.

The transformation is happening. The only question is which side of the divide you'll be on when it's complete.

Jeremy Freeman is CTO and co-founder of Allstacks, where he leads engineering and has spent the past decade helping engineering organizations understand what drives productivity.