What AI actually does to jobs — the sourced version, not the hot takes.
A data-journalism report that separates the +78M net-jobs projection from the 300M task-exposure headline. Every load-bearing figure named, dated, and footnoted.
- Why “+78 million jobs” (WEF) and “300 million exposed” (Goldman Sachs) answer different questions — and shouldn’t be compared.
- The jobs AI is creating, and the clerical and routine roles it’s replacing — with exposure broken out by sector and geography.
- Augmentation vs. replacement: why roughly half of exposed work is complemented, not eliminated (IMF).
- Charts you can trust — sourced to WEF, IMF, ILO, U.S. BLS, PwC, Goldman Sachs, and Pew, with a full bibliography.
What’s inside
Created versus displaced, resolved through one idea.
The report presents job creation and displacement as mirror images, then resolves the tension through the concept underneath all of it — augmentation. Where a point can’t be supported with a verified number, we make it qualitatively.
The Net Picture
Created vs. displaced — and why the two big numbers don’t compare.
Jobs AI Will Create
From data scientists to frontline care — where new demand actually forms.
Jobs AI Will Replace
The clerical and routine core most exposed to automation.
Automation vs. Augmentation
The central tension — most exposed work is restructured, not erased.
Exposure by Sector & Geography
Who is most affected — ~40% of global employment, ~60% in advanced economies.
Skills & the Wage Premium
What AI-fluency is worth in the labour market.
Productivity & the Global Picture
The macro effects, read against observed data.
Worker Sentiment & the DataExos View
What people feel about it — and what we conclude.
Who it’s for
For anyone who needs the real numbers, not the headlines.
Before you download
A few honest answers
Where do the numbers come from?
Named primary institutions only — the WEF, IMF, ILO, U.S. Bureau of Labor Statistics, PwC, Goldman Sachs, and Pew Research Center. Each load-bearing figure is footnoted on its page and dated, with a full bibliography at the end.
Does it predict how many jobs AI will destroy?
No single scary number — because the honest answer requires separating net headcount projections from gross task exposure. The report keeps those apart and explains why conflating them misleads.
Is it slanted optimistic or pessimistic?
Neither. It reports what the strongest available sources say, including where they disagree, and resolves the tension through the evidence on augmentation rather than a house opinion.