Many in Tech Are About to Become Unemployable — And They Can’t Even Fix Their Own Sink

...

High Pay, Low Agency

In the past, white-collar workers enjoyed both money and autonomy. A strong job market and steady demand for engineers meant they could negotiate for remote work, flexible hours, or simply walk away and land another role. Job security and mobility were baked into the deal, making the high pay feel like a genuine path to freedom.

Return-to-office mandates and shrinking flexibility signal a structural shift in leverage: by late 2024, ~90% of companies planned some form of RTO (ResumeBuilder; SHRM), and most organizations had in-office mandates (Cisco, Flex Index / ZipRecruiter) while fully remote roles contracted — see overviews here: WEF , Archie. Research links strict RTO to lower satisfaction and higher turnover, especially when autonomy is constrained (Great Place to Work synthesis).

At the same time, bargaining power is falling amid rolling layoffs and tighter hiring. Tech-sector trackers count large, continuing reductions in 2024–2025 (Layoffs.fyi / TechCrunch / TrueUp). Macro coverage echoes a broader slowdown and corporate restructurings that explicitly shift work to AI and “do more with less” targets (Washington Post). In August 2025, IT unemployment fell to 4.5% from 5.5% (CompTIA / WSJ), but postings and non-AI roles kept shrinking — volatility and uneven demand continue to erode worker agency (CIO Dive, CompTIA).

A big driver of this “high pay, low agency” regime is the AI ROI bet. Studies show AI tools can boost individual task speed (e.g., GitHub Copilot users finished a standardized task ~56% faster in a controlled experiment), which executives translate into smaller teams and fewer backfills rather than more discretion for workers (arXiv, CACM). But those gains monetize best on newer systems with small teams; they do not magically clean up legacy code or erase tribal knowledge. The result is a widening gap between output expectations and worker control — the paycheck remains, the say does not.

The Plumber vs. Engineer Paradox

Ever wondered why it takes as long to find a plumber or mechanic as it takes you to land a new engineering role? It’s not a coincidence. Plumbers and mechanics may not be guaranteed riches, but they have something most engineers lack today: control over their hours and a backlog of demand. In many U.S. metro areas, the wait time for a skilled plumber can stretch to days or even weeks, and average hourly rates have climbed above $100–$150 (HomeAdvisor / Angi data). These are signals of scarcity, not oversupply.

By contrast, software engineers are facing the opposite dynamic. The tech sector shed more than 200,000 jobs in 2024–2025, with IT unemployment hovering between 4.6% and 5.7% — above the national average — while overall U.S. unemployment stayed closer to 4.2%【101†ai-and-tech-labor.md†L29-L41】. At the same time, the pipeline keeps flooding: ~90,000 new CS graduates and ~30,000 coding bootcamp grads enter the market annually【101†ai-and-tech-labor.md†L79-L91】. Employers now face a surplus of available talent, giving them leverage to dictate terms.

The paradox is clear. For trades, scarcity translates to autonomy: plumbers can raise rates, refuse jobs, and set their own schedules. For engineers, oversupply means accepting whatever legacy work is assigned — brittle systems, under-resourced teams, mandatory overtime — because refusing could mean replacement.

This inversion of agency is structural, not cyclical. AI adoption, global labor arbitrage, and compressed team sizes mean the gap is likely to widen, not shrink. In other words: the plumber gets to choose their hours, the engineer increasingly doesn’t.

Specialization as a Trap

Hyper-specialization once looked like a ticket to stability. For decades, the advice was consistent: pick a niche, double down, and build your identity around being the expert. That worked when demand outpaced supply, but the calculus has shifted. Today, AI systems are rapidly automating the very slices of knowledge workers staked their careers on. The old advice is expiring fast — unless you’re willing to gamble that you’ll be among the 1%, maybe even the 0.01%, who not only survive but thrive in elite roles at top firms.

For everyone else, hyper-specialization has hollowed out resilience. White-collar workers outsourced almost everything outside their domain — plumbing, cars, even basic household or mechanical literacy. It wasn’t just about convenience, it was a tacit assumption: why bother learning to fix a faucet when your specialized skill in databases or front-end frameworks paid enough to hire someone else? But now, when AI begins to automate those same specialized skills, the bet looks fragile. Livelihood and identity both hinge on a narrow band of knowledge that is no longer scarce.

Trades don’t face this same fragility. A plumber’s work isn’t easily automated by AI, nor can it be offshored. They still retain agency: the ability to say no, set rates, and maintain dignity in their craft. Engineers, by contrast, find themselves in an all-in bet on a field where the odds are shifting against them. The trap of specialization is this: they can’t fire their plumber, but your boss can fire you in a heartbeat if it suits the business narrative.

The Illusion of AI ROI

Companies are cutting headcount on the assumption that AI will fill the gap. On paper, the math looks appealing: early studies show GitHub Copilot can help developers complete tasks 55–56% faster in controlled environments (Microsoft/Stanford, 2023) and internal GitHub telemetry reported a 30% increase in overall coding productivity among heavy users (GitHub blog). Executives have latched onto these figures as evidence that smaller teams can deliver the same output.

But the reality is more complex. AI scales best with small teams working on greenfield projects, not with brittle, legacy systems. As the CompTIA Tech Jobs Report (2025) shows, unemployment in IT hovers around 4.6%–5.7%, above national averages, despite record AI adoption【101†ai-and-tech-labor.md†L29-L41】. This suggests the productivity gains aren’t translating into broad employment resilience. Instead, companies are trimming staff under the belief that AI will backfill capacity, leaving remaining engineers to shoulder legacy complexity with fewer colleagues.

Research also shows that AI’s benefits drop sharply outside controlled settings. A 2023 MIT study found that generative AI sped up simple writing and code tasks for less experienced workers, but yielded negligible or no gains for complex tasks or experienced developers (Noy & Zhang, MIT). Meanwhile, CIO surveys (CIO Dive, 2024) indicate that only 11% of firms reported measurable ROI from AI pilots at scale, with most citing integration issues, hallucinations, or unanticipated maintenance costs (CIO Dive).

The structural review of tech labor confirms this: venture capital is still pouring into AI, but hiring elasticity has collapsed — investment flows no longer equate to more jobs【101†ai-and-tech-labor.md†L129-L134】. Instead, headcount is decoupled from output as productivity per developer dollar rises. Layoffs across 2024–2025, documented by TechCrunch and Layoffs.fyi, underline how widespread this recalibration has become (TechCrunch).

The net effect is that engineers end up doing more with less. AI doesn’t magically refactor millions of lines of code or erase institutional debt. It doesn’t capture the tribal knowledge lost when senior colleagues are laid off. It can scaffold snippets, but it can’t guarantee resilience at system scale. The promise of AI productivity often fails to materialize in practice, yet workers are paying the cost of corporate bets made on optimistic spreadsheets rather than grounded realities.

Work Without Ownership

Much of modern tech work amounts to cleaning up someone else’s mess: patching brittle legacy systems, carrying organizational debt, and chasing arbitrary goals set from above. It’s the same dynamic as the green pump-out boat in the marina — you’re hauling waste, but not your own. The work is necessary, but it isn’t sovereign.

For most engineers, the barriers to true ownership are everywhere. You don’t own the customer; product managers and sales teams mediate that relationship. You don’t own the infrastructure; that belongs to operations or to cloud vendors. You don’t even own the business direction, which pivots based on investor mood or executive reshuffles. At best, what most technologists “own” is a narrow slice of knowledge, and in an oversupplied market that knowledge has become a dime a dozen.

Even the code itself is rarely theirs. Employment contracts routinely assign all intellectual property to the company, regardless of who wrote it. That means the very artifact of their labor — the systems they build, the designs they author — belong entirely to the employer. The irony is stark: tradespeople like plumbers or mechanics leave behind tangible work they can point to and often build reputations on. Engineers, meanwhile, can’t claim their own code as theirs, and in a world where AI can generate boilerplate in seconds, even that claim to uniqueness is fading.

And for those who cargo cult on frameworks, they don’t even own the knowledge or the technical tradeoffs — they only know how to use a framework. The more popular the framework, the more easily it is automated by AI. All of this makes engineers replaceable.

The result is work without ownership: labor stripped of both agency and sovereignty, where contribution is reduced to cleaning someone else’s mess with no stake in the outcome and no lasting claim to the product of one’s craft.

Psychological Cost

All of these structural shifts carry a human price. The oversupply of labor, the loss of ownership, and the hollowing out of agency compound into stress and insecurity. Engineers live on a treadmill of constant learning, where every new framework or tool must be mastered on unpaid time just to keep up. Arbitrary layoffs cut across teams with little warning, reinforcing the sense that skill alone is no longer enough to secure stability. The denominator of hidden hours and mental strain grows heavier each year.

The lack of ownership intensifies that cost. Most engineers don’t feel secure because they have gone all-in — long — on technology itself. They believed the gravy train would run forever, that intellect and technical knowledge were inexhaustible moats. But moats built only on tools and syntax are eroding fast under AI. What engineers are discovering is what nearly every other trade and profession has already learned: the only truly durable moats are business knowledge, customer relationships, hard infrastructure, and taste.

When the work you do can be automated, offshored, or swept aside by a reorg, insecurity becomes the default state. Without life skills outside the terminal, without a sense of ownership, many find themselves not just replaceable but unmoored. The psychological cost isn’t just burnout — it’s the loss of confidence that one’s craft can provide a stable foundation for a dignified life.

Here is a simple question: if your tech job disappeared tomorrow, could you pick up a wrench and fix a pipe? Could you patch your own drywall or roof while you searched for your next gig? Could you make do without hiring a tradesperson for a while? If the answer is no, then you’ve bet your life and shelter on a skill set that is fast being commoditized by automation. To be clear, it’s not vanishing completely — but it is oversupplied. There aren’t enough seats for everyone. And even if you’re good, you’ll have little agency unless you are gifted, exceptional, and lucky — all three.

Historical Echoes

This isn’t new. Other fields have lived the trade of high pay for low agency, and the record is data-rich:

1) Coal miners & company towns (1880s–1930s). Many miners were paid partly in company scrip spendable only at the company store, with advances/deductions that could reach up to ~60% of pay — a financial leash that limited mobility even when wages looked acceptable (EH.Net encyclopedia; Smithsonian; Social Welfare History Project). Sources: EH.net, Smithsonian, Social Welfare History Project

2) Longshoremen before/after 1934. Hiring used to happen via the degrading “shape-up” system, where foremen chose crews on the dock. After the 1934 West Coast waterfront strike, workers won a union-run hiring hall that replaced the shape-up and instituted fair dispatch (including “low-man-out,” prioritizing those with fewer recent hours). Wages and hours improved, and autonomy rose with control over dispatch. Sources: FoundSF: The Hiring Hall, FoundSF: The Waterfront Strike, UW Libraries, Wikipedia

3) Pro athletes before free agency. Under MLB’s reserve clause, players had limited bargaining power and could be bound to teams indefinitely. In 1975, arbitrator Peter Seitz effectively ended it (Messersmith–McNally), catalyzing modern free agency. Within a few years, average player salaries jumped sharply (e.g., ~$44.7k in 1975 to ~$143.8k by 1980 in some estimates), illustrating how restoring agency (the right to choose an employer) converts directly into compensation. Sources: AP, UNCC, Westberg, Bleacher Report

4) Wall Street analysts. Compensation is high, but agency is minimal: workload and hours are dictated by clients and deal flow. The widely circulated Goldman Sachs 2021 analyst survey reported ~95–100 hours/week and ~5 hours of sleep/night; major outlets and subsequent reporting confirm the pattern of “golden handcuffs” persisting despite attempts at hour caps. Sources: Survey PDF, Guardian, FT

Takeaway. Across miners, longshoremen, athletes, and bankers, the through-line is clear: wages rise fastest when workers regain agency (hiring halls; free agency). When agency is weak (company towns; pre-free-agency leagues; junior banking), money functions as a restraint, not freedom. Tech is entering the same pattern — high compensation for a shrinking elite, low autonomy for the median worker — unless new forms of ownership and bargaining restore the ability to say no.

The Bigger Question

By now it’s clear that for most people, tech is no longer the safe harbor it once seemed. The steady salaries and career ladders have given way to oversupply, automation, and a narrowing of agency. For new entrants—and even many still employed—the goal has shifted from thriving to merely surviving. Unless you're gifted, exceptional, and lucky, the path offers less sovereignty than it once promised.

This raises the essential question: What is the dignity of labor without agency? High pay may cover the bills, but if you can't choose your hours, project focus, or claim ownership of your craft, dignity gradually erodes. History shows us that workers across industries—miners, dockworkers, bankers, athletes—eventually face this trade-off. Wages without agency are brittle. Tech is now entering that same trajectory. The real question isn't whether AI will transform the field, but whether workers can reclaim sovereignty over their labor.

Because in the end, the paycheck matters. But without agency, it leaves a lot to be desired.