Essay · March 2026

What the AI headlines miss

Sorting the real signals from the noise.

The AI headlines on any given day contradict each other. One says mass layoffs are here; another says the apocalypse was overblown. One warns workers are being deskilled by their own tools; another reports productivity has never been higher. One shows entry-level jobs collapsing; another calls the numbers a blip.

Sorting this is the discipline of anyone trying to read this moment clearly, whether you are running an organization or starting a career inside one. Three narratives dominate the current cycle: the Great AI Deskilling, AI-driven layoffs, and the collapse of entry-level work. Each contains a real signal and a layer of noise. The discipline is naming what is what.

The Great AI Deskilling.

Microsoft's New Future of Work report, released in December 2025, put the phrase into public vocabulary. AI-assisted workers look more productive on every metric; the foundation underneath may be eroding. The financial analyst no longer understands the assumptions behind the model the AI built. The junior lawyer cannot construct a legal argument from scratch because he has spent two years summarizing AI-generated briefs. "Cognitive debt" is the term that has landed. That much is real. What the panic framing misses is that some organizations are already rebuilding training around the gap. PwC launched its Learning Collective in February 2026, replacing classroom-style training with simulations, rotations, and applied client work designed to teach underlying craft before showing the AI shortcut. Some firms now require juniors to build a model or draft a brief from scratch before being allowed to run the AI version, then examine the gap between the two. The U.S. Department of Labor's $243 million AI Apprenticeship initiative, announced in April 2026, is building federal infrastructure for AI-era training paths. The pattern is the same across them: teach the underlying skill first, show the shortcut second, force the comparison that makes the skill visible. The ones who let juniors start with the shortcut will have teams that cannot function when the tools change. And the tools will change.

AI-driven layoffs.

Q1 2026 tech layoffs reached 78,557, with nearly half attributed to AI. Amazon cut fourteen thousand corporate roles; Workday cut 8.5% of its workforce; CFOs told Fortune that AI-linked cuts would be nine times higher this year than last. Some of those roles are genuinely being displaced. A meaningful fraction are something else. Over-hiring corrections that would have happened anyway now arrive wearing AI framing, because "we over-hired" is a bad reputation story and "we are being smart about AI" is a better one. The market rewards the second framing. In March 2026, Oracle announced it was cutting an estimated 20,000 to 30,000 jobs, roughly 18% of the workforce, to fund a $156 billion AI infrastructure buildout. The layoffs were real and the AI connection was real, but the mechanism was capital reallocation rather than AI replacing the people whose jobs were cut. Matt Levine has been documenting this incentive structure in his Bloomberg column for more than a year; in January 2026, Harvard Business Review ran "Companies Are Laying Off Workers Because of AI's Potential — Not Its Performance." The Klarna walkback is the canonical case: seven hundred customer-service agents replaced by AI in early 2024, quietly rehired by 2025 once quality issues surfaced. Believe the layoffs are real. Do not believe all of them are AI.

The collapse of entry-level work.

This is the sharpest data point in the set. Entry-level postings are down roughly 35% since early 2023, according to Revelio Labs. A Harvard study across sixty-two million workers by Seyed Hosseini and Guy Lichtinger found that at firms adopting generative AI, junior employment dropped 7 to 12% while senior employment kept rising. Postings that used to ask for two-to-four years of experience now ask for five. AI is absorbing the tasks that used to teach early-career workers how to do their jobs. The economic case for paying a junior to do what the AI can do has collapsed, even where the developmental case has not. The noise is in the framing as an ending. Entry-level work is being redefined. The obituary is premature. New ramps are forming around AI oversight, verification, prompt quality, human-in-the-loop work. They look different from the old ones. Operators building workforces for 2030 are designing the new ramps now. The ones who accept the "no juniors needed" default will be unable to staff senior roles in five years. If you are early in your career right now, those are the ramps to watch for. They are forming in places the old career ladder never ran.

Technology has always left skills behind. Calculators retired mental arithmetic; spreadsheets retired the handwritten ledger; GPS retired map reading. Each transition erased some capabilities and created new ones. The homesteading movement exists because so much of our daily function depends on specialized work others are doing: food production, textile manufacturing, construction. We specialize so we can add value somewhere else. AI is the next platform in that sequence. Which skills it retires is predictable. Which new skills are required to build on top of it is the harder question.

Across all three narratives, what the headlines miss is the distinction between aggregate destruction and aggregate transformation. The World Economic Forum's 2025 Future of Jobs report projects a 39% change in core skills by 2030. BCG's 2026 analysis finds AI reshaping more jobs than it replaces. MIT's Iceberg Index, released in April 2026, estimates AI could potentially replace 11.7% of U.S. jobs. The key word is "potentially." Goldman Sachs and other labor economists note that actual displacement to date is a small fraction of that potential; AI is suppressing hiring more than destroying existing jobs. None of that makes the individual losses less real for the person who has just been laid off. It does mean the story at scale is a story of change. Most roles in 2030 will look different. Most will still exist.

The losses are real. The fear is earned. The passive posture is still a choice. The people who come out of this decade ahead bring curiosity about what the work could become and the courage to redefine it from inside the role they already have, whether their own, their team's, or their children's. The work of the next decade is being designed now. The people designing it did not wait.