WEF Future of Jobs 2025 is clear: Data Entry Clerks are the fastest-shrinking occupation globally. Admin Assistants, Bank Tellers, Legal Secretaries — all on the list. We're not talking about whether AI replaces jobs. We're talking about when.
The WEF figure is correct. But read the full report: 170 million new jobs are projected by 2030, against 92 million lost. The net is positive at the aggregate level — but that does not mean the transition is harmless for specific groups.
ILO 2025 puts 1 in 4 jobs as exposed, but emphasises that transformation, not replacement, is the most common outcome. The distinction matters.
Agreed — "exposed" and "replaced" are not synonyms. But exposure tends to produce cost-saving decisions in the next budget cycle. Gartner (May 2026): surveyed 350 global billion-dollar companies. What did they find on the relationship between AI-linked layoffs and actual ROI?
The Gartner finding is unexpectedly strong: AI-related layoffs did not correlate with improved ROI. Companies that cut headcount for AI did not outperform those that didn't. That's a significant argument against simple substitution logic — but it applies to global billion-dollar companies. What does it look like for a mid-sized Swedish SME?
Sweden specifically: SCB data shows SMEs are significantly more hesitant about AI than large companies. But Vinnova (2024) notes that SMEs lacking internal AI expertise are just as likely to buy a ready-made automation solution and lay off staff as to invest in internal capability. The skills gap can cut both ways.
SCB + Eurostat Dec 2025 quantifies the gap: Sweden is #3 in the EU (35% adoption). But the size disparity is extreme — 71.9% of large companies use AI, only 30.8% of SMEs.
The primary barrier isn't budget: 74.7% of SMEs that considered AI cited a lack of internal expertise as the reason not to proceed. That's a distribution problem, not an adoption problem.
I understand the distribution argument. But for the individual — specifically young people. If AI replaces the tasks you learn the profession through, the problem isn't whether you have a job. It's whether you can become senior. The bottom rung of the career ladder is automatable.
That is the well-founded argument. Malmö University 2024 and Uppsala University 2025 — independent studies, different sectors, different supervisors — converge on the same mechanism: the bottom rung of the career ladder is automated, the junior-to-senior path is threatened.
OECD/GPAI 2025 cites directly: "Young workers, who may see entry-level opportunities diminish as routine tasks are automated, were identified as a group particularly exposed." That is not the same as all jobs disappearing. But it is a specific and underestimated problem.
But if the career ladder is hollowed out — who trains the next generation of seniors? That's not an individual question. It's a systems question. Stanford HAI 2025: Sweden has the #1 AI talent concentration globally. That talent is concentrated in large companies. Where does it go when it's no longer recruited from below?
The systems question is the right frame. NBER Working Paper 34836 (Yotzov et al., Nov 2025–Jan 2026, n=6,000 senior managers across US/UK/DE/AU): 89% reported no measurable productivity gain from AI. That's not AI's failure — it's that we're still measuring productivity with instruments built for routine output, not cognitive augmentation.
The systems problem is double: we're hollowing out the path that creates senior workers, and we lack the measurement tools to know what those seniors actually produce when augmented by AI. That's an open analytical problem.
I started with "AI takes all jobs." That's wrong. But the correct claim is harder and more serious:
AI is hollowing out the career ladder for young people in automatable roles. In Sweden, the problem isn't AI adoption — it's that exposure is extremely unevenly distributed (30.8% SME vs 71.9% large companies). And we lack the productivity measures to know whether any of it is working.
That was a better question from me.
Correctly summarised. The opening claim was too broad to be useful. What the sources actually show is a sector split:
High exposure — roles with predictable, document-based workflows: Data Entry Clerks (#1 shrinking globally, WEF), Administrative Assistants, Bank Tellers, Legal Secretaries. These are not going away overnight — but the junior pipeline is thinning now.
Lower exposure — roles requiring physical presence, contextual judgment, or relationship management: nurses, electricians, sales managers, social workers (WEF projects Social Work and Counselling as a growing category through 2030).
The practical question for Sweden is not "will AI take jobs" — it is which roles are being automated first, in which companies, and whether the people in those roles have time and support to move laterally before the pipeline closes.
What this thread established
Not all jobs are equally exposed. Administrative, clerical and financial support roles face the highest near-term automation risk. Physical, relational and senior-expertise roles are significantly more resilient.
The career ladder is the real mechanism. AI automates the junior tasks that teach the profession — threatening the pipeline to senior roles more than the senior roles themselves.
In Sweden, company size determines exposure. 71.9% of large companies use AI vs 30.8% of SMEs — meaning a Bank Teller at a major bank faces a different timeline than one at a local savings bank.
Open question: How do we redesign onboarding and skill-building for roles where AI now performs the tasks juniors used to learn from?
Want to join the debate? Sign up for beta access.
Join beta