After the Pipeline

Leverage, Not Literacy

The WEF skills chart doesn't rank what employers value. It ranks leverage — and the implicit forecast is fewer producers per orchestrator.

David H. Friedel Jr./ 2026-08-09
Subscribe
LaborAIEducation

What the WEF Skills Chart Actually Ranks

The World Economic Forum's Future of Jobs Report 20251 surveyed more than 1,000 employers representing over 14 million workers. Figure 3.4 ranks which skills employers expect to gain or lose importance between 2025 and 2030. AI and Big Data tops the list at +87. Manual dexterity sits at the bottom at −24. Programming, notably, lands at +27 — well below Creative Thinking, Systems Thinking, and Leadership.

The obvious reading is that employers value AI literacy more than coding. That reading is wrong, or at least incomplete.

The chart doesn't rank skills by importance. It ranks them by leverage.

WEFSkill.png

The Frame

Every skill on that list falls into one of two categories.

Some skills multiply output. AI and Big Data. Systems Thinking. Leadership. Teaching and Mentoring. Curiosity. A person with these skills produces more, coordinates more, or directs more than they could without them. The output scales with judgment, not headcount.

Other skills are the output. Programming, in its traditional sense. Reading, writing, mathematics. Manual dexterity. Quality control. These are skills where one person, working directly, produces one unit of work.

The chart's ranking follows this split almost perfectly. The top is dominated by leverage. The bottom is dominated by production.

That isn't employers saying "we care about AI more than code." It's employers signaling how many coders they need per orchestrator.

The chart's ranking follows a split almost perfectly. The top is dominated by leverage. The bottom is dominated by production.

Programming as the Case Study

Programming is the cleanest example because it moved categories.

Five years ago, programming was the leverage skill. A developer who could write software was multiplying the output of everyone else in the organization. Sales, operations, marketing — all of them ran on tools that programmers built.

Now programming is the skill being amplified. AI writes the code. The human specifies architecture, defines constraints, verifies output, and decides what to build. The leverage moved up a layer.

That's why Programming sits at +27 while AI and Big Data sits at +87. Not because writing code became unimportant. Because writing code became a task inside a larger job, and that larger job is what employers are hiring for.

The same compression is running through every knowledge domain on the chart. Marketing sits at +25 because AI generates the copy. Design sits at +45 because the leverage skill isn't producing mockups anymore; it's defining what should be built. Analytical thinking sits at +55 because the analysis is faster than the framing.

What Leverage Skills Have in Common

Look at the top ten:

  • AI and Big Data
  • Networks and Cybersecurity
  • Technological Literacy
  • Creative Thinking
  • Resilience, Flexibility and Agility
  • Curiosity and Lifelong Learning
  • Leadership and Social Influence
  • Talent Management
  • Analytical Thinking
  • Environmental Stewardship

None of these produce a unit of output directly. All of them determine what gets produced, by whom, under what constraints, and to what standard. They're judgment skills, coordination skills, framing skills.

That is exactly the skill profile of someone who directs an AI system rather than competing with one.

compressionimplications.jpg

The Compression Implication

Employers' ranking leverage over production tells us something specific. They aren't planning to hire more AI orchestrators to sit alongside the same number of producers. They're planning to hire fewer producers per orchestrator.

If one engineer with good systems thinking and AI fluency produces the output of ten engineers, the org chart doesn't grow to ten orchestrators. It shrinks to one. That's what compression means. Not the automation of a task. The compression of the labor required to produce a given unit of economic output.

The chart doesn't prove this is happening. But it shows employers are budgeting for it. When the skills employers most want are the skills that let one person direct the work of many, the implicit forecast is that the "many" will be fewer humans over time.

The Pattern Is Already Visible

Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen at the Stanford Digital Economy Lab published a working paper in late 2025 titled "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence."2 The chart above, reproduced in the 2026 Stanford AI Index Report,3 draws on ADP payroll data covering millions of US workers.

The pattern is unambiguous. In both software development and customer support, everyone 31 and older is growing. Everyone 26 and older is at least holding steady. And the 22-to-25 line drops — steadily, for nearly three years, in both charts. Employment for the youngest software developers fell roughly 20% from its late-2022 peak. For entry-level customer support, the drop is about 11%.

Employment for the youngest software developers fell roughly 20% from its late-2022 peak. Employment for older workers in the same jobs grew.

Entry-level cognitive work is compressing first. That's not a forecast. That's payroll data.

The pattern matches what the skills chart predicts. Employers are ranking leverage skills at the top because they're hiring for leverage. They're deprioritizing production skills because production is what AI now does. The 22-to-25 cohort in software isn't disappearing because young people stopped learning to code. They're disappearing because the entry-level task — writing routine code under supervision — is the task AI is best at.

The 2026 AI Index Report3 notes that one-third of surveyed employers expect further workforce reductions over the coming year. The chart isn't a snapshot of a completed shift. It's the beginning of one.

The Pipeline Problem

This creates a tension the WEF chart doesn't name.

Employers want senior people with leverage skills. Systems thinking, architecture, judgment, the ability to direct AI systems. Those skills don't come from a course. They come from years of doing production work under senior supervision — the exact stage of the career that just compressed.

The top of the skills chart demands orchestrators. The bottom of the Brynjolfsson chart shows the pipeline that produces them shrinking. Employers are optimizing for a workforce whose supply mechanism they're simultaneously eliminating.

This is where the tractors-and-spreadsheets rebuttal fails hardest. Historical automation displaced specific tasks. Workers reallocated to adjacent work at similar career stages. AI is displacing specific career stages across every knowledge domain simultaneously. There is no adjacent stage to reallocate to. You can't skip the entry level and become a senior orchestrator. The industries hiring at the top aren't hiring at the bottom, and the industries that used to hire at the bottom aren't hiring at all.

73f91c17-71c8-470f-ad31-1a4ab25da4a1-2026-08-09.png

What to Watch

The macroeconomic signature of this pattern is measurable and already visible.

Revenue per employee at the frontier AI labs is an order of magnitude above traditional software companies at similar revenue. Epoch AI's analysis4 puts Anthropic and OpenAI at roughly $14M and $6.5M in revenue per employee — higher than any tech company on the Forbes Global 2000.

S&P 500 operating margins5 closed 2025 at 13.16%, roughly 19% above the long-term average of 11.04%. Q1 2026 blended net margins hit 13.4%6 — the highest on record. Margin expansion has been running consistently since late 2023, powered largely by tech.

Labor's share of national income has been drifting down since the early 2000s and shows no sign of reversing. Bureau of Labor Statistics data shows labor's share fell from 63.3% in 2000 to 56.7% in 2016,7 with three-quarters of the entire post-1947 decline occurring in that window. It has remained near those lows since.

Revenue per employee at the frontier AI labs is an order of magnitude above traditional software companies at similar revenue.

The occupational signature is what Brynjolfsson is measuring. If compression is real, the 22-to-25 line keeps falling and the 26-to-30 line follows it. If the pipeline problem is real, the 31-plus lines eventually plateau — because you can't grow the senior cohort past what the junior cohort feeds into it, and the junior cohort is shrinking.

The WEF chart answers a smaller question than "more jobs or fewer jobs." It answers what employers value. But what they value is telling, and what they're already doing is more telling still. They're not preparing to hire more people who can produce. They're already hiring fewer of them.

That is the compression thesis, written in the language of a payroll.

Footnotes

  1. World Economic Forum, Future of Jobs Report 2025, January 2025. — The survey this piece reads against the grain: more than 1,000 employers representing over 14 million workers, and Figure 3.4 ranking which skills gain or lose importance between 2025 and 2030. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
  2. Brynjolfsson, Erik; Chandar, Bharat; Chen, Ruyu. "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence." Stanford Digital Economy Lab, November 2025 draft. — The payroll evidence that turns the chart's implication into a measurement: ADP data across millions of U.S. workers showing entry-level employment falling in software and customer support while older cohorts hold or grow. https://digitaleconomy.stanford.edu/app/uploads/2025/11/CanariesintheCoalMine_Nov25.pdf
  3. Stanford Institute for Human-Centered AI, 2026 AI Index Report, Economy chapter, April 2026. — Reproduces the Brynjolfsson cohort chart and carries the forward-looking figure cited here — one-third of surveyed employers expecting further workforce reductions in the coming year. https://hai.stanford.edu/ai-index/2026-ai-index-report/economy
  4. Emberson, Luke. "Anthropic and OpenAI earn more revenue per employee than major public tech companies." Epoch AI, 2026. — The clearest single measure of leverage as an operating fact rather than a forecast: roughly $14M and $6.5M in revenue per employee, above any tech company on the Forbes Global 2000. https://epoch.ai/data-insights/revenue-per-employee-ai-companies
  5. S&P 500 Operating Margin historical data. GuruFocus, S&P Dow Jones Indices. — Source for the 2025 close at 13.16% against a long-term average of 11.04% — the margin side of the same compression the payroll data shows on the labor side. https://www.gurufocus.com/economic_indicators/4226/sp-500-operating-margin
  6. FactSet blended net profit margin data, Q1 2026, via MoneyShow / Seeking Alpha, April 29, 2026. — Record blended net margins of 13.4%, with the expansion running since late 2023 and driven largely by technology. https://seekingalpha.com/article/4895675-chart-of-day-expanding-margins-help-fuel-s-and-p-run
  7. McKinsey Global Institute, A new look at the declining labor share of income in the United States. — The long-run backdrop: labor's share falling from 63.3% in 2000 to 56.7% in 2016, with three-quarters of the entire post-1947 decline concentrated in that window. https://www.mckinsey.com/featured-insights/employment-and-growth/a-new-look-at-the-declining-labor-share-of-income-in-the-united-states
Back to the Journal