The year the narrative snapped

If you’ve seen the headline “>100,000 Big Tech layoffs in 2025,” take a breath. The broader tech industry count sits closer to ~80.8k YTD, versus 152.9k in 2024 and 264.2k in 2023—a downshift, not an extinction event. Cost discipline, higher rates, and AI-efficiency programs are all in the mix; a pure “AI shock” is hard to isolate in the data. Layoffs.fyi

Zooming out, credible institutions keep stressing exposure, not inevitability. The IMF estimates about 60% of jobs in advanced economies are exposed to AI; roughly half of those may face downward pressure on tasks, wages, or demand, while the rest are augmented. Globally, exposure is closer to 40%. “Affected” ≠ “replaced.” IMF+1

Callout: The smartest read of 2025 so far: AI is a force multiplier for organizations—and a force reshaper for job ladders.

Who is affected—and how?

The OECD’s 2025 analyses add a geographic and equity lens: regions once seen as “automation-safe” (think capital regions heavy on knowledge work) are often the most exposed to generative AI. Exposure and displacement risks also skew higher for workers with lower formal education and for migrants. Translation: the fairness problem is spatial and social, not just technical. OECD+2OECD+2

Meanwhile, employers are planning around this reality. The WEF’s Future of Jobs 2025 finds ~40% of employers anticipate reductions where tasks are automatable, even as demand spikes for data, AI, and FinTech roles. Net effects depend on rollout pace and reskilling depth. World Economic ForumWeltwirtschaftsforum Berichte

The productivity surge is real. Hiring? Not necessarily.

Field evidence keeps stacking up:

  • Customer support: A staggered rollout of a gen-AI assistant increased issue resolution ~14% on average, with the biggest gains for novices. NBER+2NBER+2
  • Software dev: In a controlled experiment, developers with GitHub Copilot finished tasks ~55.8% faster. MicrosoftarXiv
  • Public sector: A cross-government UK trial reports ~26 minutes saved per day per civil servant using Microsoft 365 Copilot. Gouvernement Großbritannien+1

But productivity ≠ headcount growth. Companies are using AI to do more with the same—or fewer—people. Remember the high-profile signals:

  • Klarna: Its AI assistant handled two-thirds of service chats and claimed ~700 FTE equivalent output with higher first-contact resolution. Klarna
  • IBM: The CEO projected ~30% fewer back-office roles over five years (roughly 7,800 positions) due to AI/automation. BloombergAl Jazeera
  • BT Group: Plans to cut 40k–55k roles by 2030, with ~10k explicitly tied to AI and automation—an example of sector-wide restructuring, not just “tech.” Der GuardianForbes

The bottom line: teams get faster; ladders get shorter. AI tends to boost experienced workers first—and compress the entry rungs beneath them.

The quiet culling of entry-level jobs

If you’re early-career, 2025 feels tighter. Tech job postings on Indeed are ~36% below Feb 2020, and the freeze has persisted since mid-2023. Senior roles remain, but those “learn-on-the-job” openings? Thinner. Indeed Hiring Lab

Freelance markets show measurable substitution, too. In Upwork categories highly exposed to gen-AI, monthly jobs fell ~2% and earnings ~5.2% post-GenAI—small but statistically meaningful. That’s the routine white-collar slice getting automated at the margin. American Economic AssociationSSRNBrookings Institution

This is the structural tension: AI raises the ceiling on output and lowers the need for the bottom rungs where people historically learned and advanced.

Guardrails are arriving—do they help?

Regulators are building speed bumps where the risks are clearest: hiring, monitoring, and evaluation.

  • U.S. Department of Labor (2024): AI & Worker Well-Being Principles and Best Practices—guidance to reduce harm, bias, and job-quality erosion. DOLprivacysecurityacademy.com
  • NYC Local Law 144: Bias audits and notices required for Automated Employment Decision Tools (AEDTs) used in the city—audits refreshed annually. NYC+1
  • EU AI Act: Employment and worker-management systems are presumed high-risk (Annex III), with staged obligations through 2025–2026+. HR tech will need documented risk controls, transparency, and human oversight. EU Artificial Intelligence Act+1Europäisches ParlamentDigitale Strategie Europas

These rules won’t “create jobs,” but they can dampen reckless deployment—especially in screening and performance surveillance—while raising the standard for fairness and explainability.

Beware neat numbers: exposure isn’t destiny

Claims like “20% of jobs gone in five years” are attention-grabbing—but they’re not consensus forecasts. The IMF and WEF emphasize exposure and employer intentions, not guaranteed losses. Think of AI as shifting task mixes, wage pressures, and hiring thresholds—with outcomes mediated by policy and training. IMFWorld Economic Forum

Quote to hold onto: “Affected doesn’t mean replaced.” Exposure data tells you where to invest in people, not whom to discard.

What leaders should do next (practical playbook)

1) Protect the pipeline. Make space for structured apprenticeship-style roles—pair juniors with AI tools and seniors. Require documented skill transfer, not just task offloading.

2) Measure augmentation, not vibes. Track time saved, defect rates, customer outcomes, and novice lift (the NBER call-center study’s biggest effect). Tie AI ROI to team outcomes, not headcount cuts alone. NBER

3) De-risk hiring AI. If you deploy ranking/screening tools, align to NYC AEDT audit standards now; it’s a good proxy for EU high-risk HR expectations later. Build explainability and adverse-impact monitoring into your model lifecycle. NYCEU Artificial Intelligence Act

4) Train for the bottlenecks. Prioritize skills with durable demand—data, ML ops, cybersecurity, cloud, product judgment—and fund AI-assisted upskilling for internal mobility.

5) Communicate the social contract. If productivity gains reallocate work, say how. Share the plan: reskilling budgets, internal transfers, and where AI won’t be used (e.g., termination decisions).

Why This Matters

The AI and jobs debate is not a binary of job-killer vs. job-creator. It’s a distribution story—who benefits, who absorbs the shock, and where ladders get rebuilt. The data shows real productivity gains, entry-level compression, and growing policy guardrails around HR systems. The future we get hinges on how deliberately we implement AI: pairing augmentation with apprenticeships, audits with transparency, and efficiency with a genuine plan for human growth. NBERMicrosoftIndeed Hiring LabNYCEU Artificial Intelligence Act

Internal reads (Vastkind)

  • The AI Productivity Paradox: Why Output Soars but Hiring Stalls (Vastkind)
  • Designing Fair Hiring Pipelines in the Age of AEDTs (Vastkind)
  • From Pilots to Policy: Building an AI Governance Playbook (Vastkind)

External references (authority)

  • IMF — Gen-AI: Artificial Intelligence and the Future of Work (exposure: 60% in advanced economies; global ~40%). IMF
  • OECD — Emerging Divides in the Transition to AI and Effects of GenAI on Productivity, Innovation & Entrepreneurship. OECD+1
  • WEF — Future of Jobs 2025 (employer intentions; 40% anticipate reductions where tasks are automatable). World Economic ForumWeltwirtschaftsforum Berichte
  • NBER — Generative AI at Work (contact-center productivity +14%; strongest for novices). NBER
  • Microsoft Research / arXiv — The Impact of AI on Developer Productivity (Copilot ~55.8% faster in controlled task). MicrosoftarXiv
  • UK Government — Microsoft 365 Copilot Experiment: Cross-Government Findings (26 minutes/day saved). Gouvernement Großbritannien
  • AEA / SSRN — Short-Term Effects of Generative AI on Upwork (-2% jobs, -5.2% earnings in exposed tasks). American Economic AssociationSSRN
  • Layoffs.fyi — Tech layoffs tracker (context across 2023–2025). Layoffs.fyi
  • NYC DCWP — AEDT bias-audit and notice requirements. NYC
  • EU AI Act — High-risk classification for HR systems; implementation timelines. EU Artificial Intelligence ActEuropäisches Parlament

Conclusion

2025 is the year AI’s labor story got specific. The scare stats don’t hold up, but neither does the fairy tale. We’re watching measurable productivity, tighter entry doors, and the first serious guardrails for HR tech. The leaders who win will be those who protect the talent pipeline while codifying fair, explainable AI into hiring and performance systems.

Your move: If this is your battlefield in 2025, subscribe to Vastkind’s Field Notes for monthly, data-driven briefings—or DM us to build a live AI-and-Jobs tracker (cases, studies, policy timeline) for your org.