Recent Goldman Sachs data shows that unemployment among 20 to 30 year old tech workers has jumped nearly 3 percentage points since early 2025, over four times the national average. PwC announces that they'd be reducing entry-level hiring by up to 39% over the next 3 years. Cursor reached $100M ARR with just 20 employees and Replit goes from $10M to $100M ARR in just 5.5 months.
Something fundamental is shifting in our industry, and the data tells a story that's both complex and concerning than the typical "AI will create more jobs than it destroys" narrative we keep hearing.
The Two Different Stories
Let me start with what seems like good news. According to PwC's analysis of nearly a billion job advertisements, productivity in AI-exposed industries has quadrupled from 7% growth (2018-2022) to 27% growth (2022-2024). Companies using AI are seeing revenue per employee grow 3x faster than those that aren't. Workers with AI skills command a 56% wage premium, up from 25% just a year ago.
But here's where it gets interesting, and by interesting I mean troubling if you're trying to break into tech. The ICT (Information and Communication Technology) industry's share of total job postings has nearly halved over the past 12 years. The absolute number of tech jobs continues to grow, but as a percentage of all jobs available, tech is shrinking. We're becoming more productive, more valuable, but proportionally smaller.
Goldman Sachs puts an even finer point on it: tech sector employment peaked in November 2022, the exact month OpenAI launched ChatGPT. Since then, it's declined steadily below pre-pandemic trends. This marks the end of over 20 years of linear growth in tech employment.
The Great Restructuring
What we're witnessing isn't the mass unemployment event that doomsayers predicted. It's something more subtle and perhaps more profound: a complete restructuring of how work gets done and who gets to do it.
Microsoft's AI platform is building systems that learn and improve with every interaction. They're moving from org charts to work charts, from hierarchical reporting structures to task-oriented team formation where humans and AI agents collaborate fluidly.
This shift from static structures to dynamic systems isn't unique to Microsoft. It's becoming the industry standard. Goldman Sachs has "hired" Devin, an AI software engineer, deploying hundreds of instances initially with plans to scale to thousands. These AI agents work alongside their 12,000 human developers, delivering what they expect to be a 3-4x productivity increase. One of Wall Street's most prestigious firms, known for its selective hiring and analyst programs, is now onboarding AI agents with the same seriousness they once reserved for Ivy League graduates.
Across the industry, we're seeing this same pattern play out. Klarna reduced its customer service team from 700 to 70 while improving resolution times. Stability AI can generate more images daily than all human artists combined produced in the 20th century, with a team smaller than a typical McDonald's franchise. Even traditional enterprises like Walmart and JPMorgan are deploying thousands of AI agents, fundamentally reorganizing how work flows through their organizations.
The implication? Companies need fewer people to do the same work, but those people need to be significantly more skilled. The entry-level positions that traditionally served as the industry's training ground are disappearing.
Winners, Losers, and the Uncomfortable Middle
Looking at the data, three distinct groups are emerging in this transformation.
First, the clear winners: experienced professionals who can work with AI systems. These are the people seeing those 56% wage premiums. They're the mid to senior engineers who can now operate with the leverage of an entire team. They're the product managers who can prototype their own ideas. They're the full-stack builders that companies like Cursor and Midjourney represent, generating hundreds of millions in revenue with small teams.
Second, the potential losers: entry-level workers trying to break into tech. When even PwC, a company built on pyramid-shaped hiring models, is cutting entry-level positions by a third, you know something fundamental has changed. The traditional path of joining a big tech company as a junior developer, learning on the job, and climbing the ladder? That path is narrowing to the point of near-closure.
But there's a third group that's perhaps most interesting: the specialized vertical experts. Y Combinator makes a compelling case that we're about to see hundreds of vertical AI agent companies, each worth billions, automating specific industry workflows. Take Candid Health, founded by someone who watched their parents navigate medical billing nightmares, now processing billions in claims with AI-powered operations. Or Harvey AI, built by securities lawyers who understood exactly which legal workflows could be automated, now valued at $1.5 billion. Or even smaller players like Norm AI, which turned compliance expertise into an AI platform that helps companies navigate regulatory requirements.
These people are industry insiders who've lived the problems they're solving. A former restaurant manager building scheduling AI that actually understands split shifts and labor laws. A construction project manager automating bid analysis with deep knowledge of material costs and timeline dependencies. They have something that can't be easily replicated: years of accumulated context about where their industries actually break down.
The Productivity Paradox
If productivity is increasing this dramatically, where does the value accrue?
McKinsey estimates AI could add $4.4 trillion in global economic value which is at $109.7 trillion for 2024. But their research also shows that only 1% of companies have achieved AI maturity, despite 92% planning to increase investments. There's a massive gap between AI's potential and current reality. Companies that successfully integrate AI will see margins expand dramatically. Those that don't will find themselves unable to compete. We're already seeing this play out. AI leaders outperform laggards by 2-6 times on total shareholder returns, and that gap increased 60% over just three years.
But will this lead to a few mega-winners or spark new competition? The data suggests both. On one hand, the capital requirements for training large models create natural monopolies. On the other, Nathan Lambert, a machine learning researcher, says that the shift from pre-training to post-training means smaller players can compete by specializing. You don't need to train a foundation model; you need to fine-tune existing ones for specific use cases.
The low headcount, high revenue companies we're seeing suggest that barriers to entry are simultaneously rising and falling. Rising because you need AI expertise. Falling because you need far less capital to build and scale a product.
The Skills Earthquake
PwC found that skills requirements are changing 66% faster in AI-exposed occupations compared to historical rates. What you learned in a four-year computer science degree might be obsolete by graduation.
This is driving a fascinating shift in hiring. Formal degree requirements are falling, especially in AI-exposed jobs. In ICT roles, degree requirements dropped from 66% to 59% for AI-augmented positions. The industry is moving toward skills-based hiring, prioritizing practical AI competency over formal qualifications.
Companies want individuals who can both code and translate business problems into AI solutions. They're the ones who understand both the capabilities, limitations of these systems and work across traditional disciplinary boundaries.
The Wage Gap Widens
The 56% wage premium for AI skills tells only part of the story. What we're really seeing is a bifurcation of the labor market. McKinsey's research shows that employees are adopting AI far faster than their managers realize. 13% of employees already use AI for 30% or more of their daily work, but only 4% of C-suite leaders believe this is happening. The gap between those who get it and those who don't is widening at every level of organizations.
Young professionals face a particularly cruel irony. They're the most comfortable with AI (62% of millennials report high AI expertise) but also the most likely to be displaced by it. They have the skills but lack the experience that makes those skills valuable. It's a catch-22: you need experience to add value in an AI world, but the entry-level jobs that provide that experience are disappearing.
What Happens to the IT Industry?
Here's what I think happens next. The IT industry becomes smaller but more valuable. Fewer people will work in tech relative to other industries, but those who do will be exceptionally well-compensated. We're moving from an industry that employed masses at good wages to one that employs fewer people at exceptional wages.
Big companies will get bigger in revenue, market cap, and influence while actually shrinking their workforce. Microsoft's productivity gains, Goldman's AI employees, PwC's restructuring, they all point in the same direction: massive organizations run by small teams.
But these same forces will enable new competition. When 20 people can build a $100M business, when vertical AI agents can automate entire workflows, when the marginal cost of software approaches zero, the traditional moats of big tech companies start to look less defensible. Network effects matter less when AI agents can integrate anything with anything. Economies of scale matter less when small teams can leverage infinite compute.
We'll see both consolidation and fragmentation. Consolidation at the platform layer (the companies training foundation models, providing infrastructure) and fragmentation at the application layer (thousands of vertical AI companies solving specific problems).
Looking Forward
The next few years will likely separate those who use AI thoughtfully from those who use it reactively. Organizations implementing AI successfully gain elastic engineering capacity, compressed timelines, and operational leverage that becomes a sustainable competitive advantage. For individuals, the path forward requires embracing continuous learning as a core competency. The 66% acceleration in skills change means what you know today will be table stakes tomorrow. This has always been true in tech but this time the timelines are compressed. We used to have years to adapt to new paradigms. Now we have months, maybe weeks.
The IT industry isn't dying. It's shedding its skin. What emerges will be leaner, more powerful, and fundamentally different from what came before. Those who thrive will be those who understand it deeply enough to navigate it. The question is whether AI will preserve what made tech special in the first place: the belief that anyone with curiosity and determination could build something that matters.