I had three recruiting conversations last week that shook me.
First call: A senior backend engineer with 12 years at FAANG companies, laid off in January, still unemployed in April. “I’ve applied to 247 positions,” he told me. “I’m getting rejections for roles I would’ve been overqualified for two years ago.”
Second call: A hiring manager at a Series C startup, desperate to fill two AI/ML engineering roles that have been open since October. “We’ve interviewed 40 candidates. None have the right mix of traditional engineering AND AI fluency. We’re paying 40% above market and still can’t close.”
Third call: A mid-career engineer at my company, asking in our 1:1 if she should be worried. “I see the AI tools. I’m trying to learn. But I have a mortgage and two kids and I’m terrified I’m on the wrong side of something I don’t fully understand.”
Three conversations. One pattern. The tech labor market didn’t shift—it split in two.
The Numbers Demand Our Attention
Tech unemployment hit 5.8% in early 2026—the highest level since the 2001 dot-com bust—while overall U.S. unemployment sits at 4.1%. We’re an outlier, and not in a good way.
The industry has already shed 55,775 jobs across 166 companies in just the first 74 days of this year. At this pace, we’re projected to hit 264,730 layoffs by December, surpassing 2025’s 245,000 and marking this as the worst year for tech employment in a generation.
Here’s what’s different this time: 20.4% of these layoffs are explicitly attributed to AI and automation. Out of 45,363 confirmed global tech layoffs through early March, companies cited AI as a factor in 9,238 cuts. That’s up from less than 8% in 2025. Companies aren’t just restructuring—they’re restructuring BECAUSE of AI.
And here’s the human cost buried in those numbers: median time to re-employment has increased from 3.2 months in 2024 to 4.7 months in early 2026. That’s nearly 50% longer. Five months unemployed means depleted savings, eroded confidence, decaying skills, and mounting desperation.
It’s Not a Shift—It’s a Bifurcation
The narrative I keep hearing is that the labor market is “shifting” toward AI skills. That’s too gentle. What’s happening is a split into two distinct labor markets:
Side A: Engineers with AI fluency—prompt engineering, AI governance, LLM fine-tuning, agent orchestration, model evaluation. These roles are seeing 777% growth in demand for prompt engineering alone, 1,257% for AI governance. Companies are paying 40% premiums and still struggling to fill positions. Time to hire? Months, even for well-funded startups.
Side B: Traditional engineers competing for a shrinking pool of roles that don’t require AI specialization. These are the folks with 12 years of backend experience sending out 247 applications. These are the mid-career engineers asking if they should be worried. These are the 55,775 laid off so far this year.
The gap between these two markets isn’t just a skills difference—it’s an employability chasm. And here’s the brutal math: only 16% of workers have high AI readiness (defined as the skills, fluency, and operational context to work effectively alongside AI tools), yet 75% of companies are adopting AI while only 35% of their talent has received AI training in the last year.
We have a structural mismatch at scale.
The Skills Debt Is Compounding
I’m not talking about “learning to use ChatGPT.” I’m talking about the delta between what the market is hiring for and what most engineers currently know:
- Prompt Engineering demand reached 121,000 job postings in H2 2025 alone
- AI Governance roles grew 1,257% across 18 months
- LLM Fine-Tuning expertise grew 1,161%
- Yet 20% of organizations say AI model and application development skills are the most difficult to hire for
This isn’t a training gap you close with a Coursera course and a weekend project. This is a fundamental reorientation of how we think about engineering work—from writing code to orchestrating systems of AI agents, reusable components, and external services.
And here’s what keeps me up at night: while 75% of companies are adopting AI, only 35% of their existing employees have been trained. That means 40% of the workforce at AI-adopting companies is being left behind by their own employers. Not because they’re incapable, but because we’re not investing in the transition.
The Human Cost of 4.7 Months
Let’s sit with that re-employment number for a moment. 4.7 months. Up from 3.2 months just two years ago.
That’s not just a statistic—that’s someone’s mortgage payment, health insurance, child’s tuition, retirement savings, mental health. That’s 20 weeks of job applications, coding challenges, system design interviews, behavioral rounds, and rejections. That’s 140 days of wondering if you’re unhireable at 45, or 38, or 52.
And the longer someone is unemployed, the harder it gets. Skills decay. Confidence erodes. Desperation sets in, and companies can smell it. The tech industry’s ageism problem compounds the AI skills problem—if you’re 50 and don’t have AI fluency, you’re facing a double barrier.
We’re creating a class of permanently displaced tech workers who are too experienced (read: expensive) for companies that want to hire junior AI talent, and too under-skilled in AI for roles that require it. The middle is disappearing.
The Leadership Question We’re Not Asking
Here’s what I want to know from other leaders in this community:
What are we actually DOING to help our existing teams cross this chasm?
Are we:
- Investing in structured AI upskilling programs, or just pointing people to Coursera?
- Creating internal AI apprenticeships where senior engineers learn by doing, not by watching videos?
- Measuring AI productivity gains honestly, or justifying layoffs with imagined efficiencies?
- Having transparent conversations about which roles are at risk and why, or leaving people to guess?
- Treating AI training as a retention strategy or as performance theater?
Because if we’re being honest, most companies are doing the bare minimum. We’re offering Copilot licenses and calling it a training program. We’re encouraging “self-directed learning” because we don’t want to allocate the time or budget. We’re measuring “AI adoption” by counting tool licenses, not by tracking actual capability building.
And then we’re surprised when we have to hire externally for AI roles because our existing teams aren’t ready.
Which Side Are You On—And What Are You Doing About It?
This isn’t a temporary disruption. This is a structural labor market bifurcation that will define careers for the next decade.
If you’re a senior engineer who hasn’t touched an AI tool in anger, you’re on Side B—and the distance to Side A is growing every month.
If you’re a manager or director, your team is watching you to see if you’re investing in their future or managing their decline.
If you’re a VP or CTO, your budget decisions right now—upskilling vs. layoffs, internal development vs. external hires—will determine whether you retain institutional knowledge or start from scratch.
I don’t have all the answers. But I know this: the engineers asking “which side am I on?” are the ones paying attention. The ones still assuming this will blow over are the ones who will be blindsided when their severance package includes a Coursera subscription and a LinkedIn Premium trial.
So I’ll ask the hard question: Which side of the AI labor market split are YOU on—and what are you doing this week, not someday, to make sure you’re prepared?
Because 4.7 months is a long time to figure it out when you’re already on the market.
Sources: