3.7 Million. That's How Many Clerical Jobs the WEF Says Disappear by 2030.
The World Economic Forum's Future of Jobs Report 2025, published in January of that year, runs to 290 pages and contains enough projections to keep a labor economist busy for a semester. But one number keeps surfacing in policy briefings and company restructuring announcements: 3.7 million. That is the net displacement figure the WEF assigns to "clerical and secretarial workers" across the 22 industry clusters and 55 economies it surveyed. Not gross job losses — net. Meaning 3.7 million more positions eliminated than created, in a single occupational category, within five years.
To understand why that number is both alarming and somewhat slippery, you have to read the methodology section that most people skip.
What the WEF Actually Surveyed
The WEF did not count job postings or run payroll records. It surveyed roughly 1,000 employers, collectively representing about 14 million workers across its member economies, and asked HR executives to estimate the share of roles in their organizations they expected to grow, shrink, or disappear entirely by 2030. The 3.7 million figure is then extrapolated from those employer-reported intentions to the broader labor force using employment weight data from the International Labour Organization.
That methodology has an obvious weakness: it measures what employers plan, not what they will do. Corporate intentions surveys are notoriously optimistic about efficiency and notoriously vague about timing. The WEF's own previous iteration — the 2023 edition — projected 83 million gross job losses globally across all sectors over its five-year window, a figure that looked dramatic in press coverage but became difficult to verify against actual BLS or Eurostat data because "global" and "gross" are doing enormous definitional work.
Still, the clerical category is worth isolating. Unlike, say, "data entry clerks" (a narrow occupation the Bureau of Labor Statistics has tracked declining for a decade), "clerical and secretarial" in WEF taxonomy is broad. It covers administrative assistants, bookkeeping clerks, payroll processors, office managers, and several sub-roles that handle document routing and scheduling. These are jobs where a substantial portion of daily task time — the WEF's own time-use analysis puts it at 51 percent for this category — involves processing, organizing, or transmitting structured information. That is precisely what large language models have gotten good at.
The Goldman Sachs Number That Sits Next to It
In March 2023, Goldman Sachs Economics published a research note titled "The Potentially Large Effects of Artificial Intelligence on Economic Growth," authored by Joseph Briggs and Devesh Kodnani. Their headline figure — 300 million jobs globally exposed to automation — is the one that traveled furthest on social media. Less quoted was their sector breakdown, which estimated that roughly 46 percent of administrative and office-support tasks in the United States are "potentially automatable" with current generation AI tools.
Goldman was careful to say "exposed" rather than "eliminated." Exposure means the task composition of a job overlaps significantly with AI capability — it does not mean the job vanishes. A claims processor whose role is 60 percent data entry and 40 percent client judgment calls is "exposed" at 60 percent, but whether the employer eliminates the role, restructures it, or simply squeezes more throughput from fewer people depends on a dozen factors Goldman explicitly did not model: union contracts, regulatory requirements, manager inertia, the actual cost of enterprise AI deployment.
When you hold the Goldman "exposure" framing next to the WEF "net displacement" figure, they are measuring orthogonal things. Goldman tells you how much of a job could be automated given current technology. The WEF tells you how many jobs employers intend to cut net of new hires. Neither tells you what will actually appear on payroll records in December 2030.
What BLS Data Shows Right Now
The Bureau of Labor Statistics Occupational Employment and Wage Statistics program, updated through May 2025, shows that "secretaries and administrative assistants, except legal, medical, and executive" numbered approximately 2.1 million in the United States. That is already down from roughly 2.5 million in the 2019 count — a 16 percent decline over six years that predates the widespread deployment of LLMs by several years, driven partly by cloud collaboration tools and partly by flatter organizational structures post-pandemic.
The WEF's 3.7 million figure is global across 55 economies, so it cannot be compared directly to the BLS occupational slice. But the directional alignment is real: clerical work was already shrinking before GPT-4 launched, and the technology curve since late 2022 has accelerated the task automation that was previously theoretical.
What the Number Misses
The displacement figure says nothing about where the workers go, which is the more consequential question. The WEF's same report projects 69 million gross job gains globally by 2030, most of them in green transition roles, care economy positions, and technology operations. The uncomfortable arithmetic is that the workers being displaced from clerical roles and the workers who would fill, say, solar panel installation or AI training data curation jobs are not interchangeable populations. Geographic mismatch, credential mismatch, and age mismatch do work that aggregate numbers conceal.
The Stanford AI Index 2025, published by the Stanford Institute for Human-Centered AI, noted that occupational transition rates in the United States have historically taken eight to twelve years to equilibrate after major automation shocks — longer than the WEF's five-year projection window. That gap is not a flaw in the WEF model; it is simply outside the scope of what employer intention surveys can capture.
So What Does 3.7 Million Actually Predict?
Taken seriously but not literally, the number predicts that clerical work is in the fastest-moving segment of the current displacement cycle, that the technology required to automate its core tasks already exists and is commercially deployed, and that the employers closest to those roles believe the headcount reduction will be substantial within the current business planning horizon. Whether the final tally is 2 million or 5 million by 2030, the people currently holding those roles have less runway than the aggregate number suggests, because aggregates never displace anyone — specific employers making specific decisions do, and those decisions are already underway.
PROTECT YOUR INCOME NOW
If AI automation threatens your job, preparation is the only honest answer: know your risk, build a runway, and reskill before you have to. The Clankers 90-day programme structures exactly that — free, which is exactly why today is the right time to start.
Start the 90-Day Programme — Free →GET THE CLANKERS BRIEF — FREE
One email every Monday: the week's AI displacement moves that matter, and what to do about them. No spam. Unsubscribe anytime.