"Is AI replacing jobs?" has become more of a social and political question than an economic one. That means most commentary is catastrophising or dismissing based on ideology rather than data.
Here is what the data actually shows in mid-2026.
Global Unemployment Has Not Increased
If AI is replacing jobs at significant scale, unemployment should be rising. It is not.
Global unemployment trend (ILO data): - 2023: 5.8% - 2024: 5.7% - 2025: 5.8% - 2026 Q1: 5.7%
In the US: 3.9% in Q1 2026, down from 4.0% in 2025. Job creation remains positive at ~200,000 new jobs per month.
What this tells us: If AI was causing net job destruction at scale, unemployment would rise. It has not.
What this does not tell us: Jobs are not disappearing everywhere uniformly. Job destruction in some sectors is being offset by job creation in others. The compositional shift is real even though aggregate numbers hide it.
Job Composition Is Shifting
Where measurable change IS visible:
Within-sector displacement: The junior analyst role is declining; the senior AI-augmented analyst role is growing. Tier 1 customer service is shrinking; complex case management is growing.
Sector growth rates diverging: - ML engineering: +22% year-over-year - Nursing: +8% year-over-year - Software engineering: +12% year-over-year - Customer service: -3% year-over-year - Clerical support: -8% year-over-year - Retail sales: -5% year-over-year
Compensation in automation-pressure roles stagnating: - Customer service representative: -2.5% salary from 2024-2026 - Data entry specialist: -3.8% salary - Junior financial analyst: -1.2% salary
- ML Engineer: +18% salary 2024-2026
- Senior software engineer: +9% salary
- Healthcare specialties: +5-7% salary
AI Adoption Speed vs. Historical Baselines
Useful comparison: how fast is AI moving vs. previous technology transitions?
ATM adoption (1990s-2000s): - Potential: replace 50-70% of bank teller jobs - Actual timeline: 20 years to reach that level - Outcome: Bank teller employment actually grew 5% despite ATMs (branch openings exceeded automation impact)
E-commerce adoption (2000s-2010s): - 10-year window to see major retail employment displacement - Peak annual decline: -2.5% (during 2008-2009 recession)
Current AI adoption (2023-2026 forward): - Potential: Automate 45-60% of customer service / certain admin roles - Actual timeline so far: 2-3 years to 45% deflection to AI - Annual decline in affected categories: -3% to -8%
Assessment: AI adoption is moving faster than ATM or e-commerce, but not shockingly faster. At current speed, most automatable work would take 5-10 years to displace, not 2-3 years.
Concentration of Displacement Risk
Displacement is not uniform. It is concentrated.
By role: - Highest risk (>70%): Data entry, transcription, routine customer service, bookkeeping, routine QA - Medium risk (40-70%): Junior analyst, junior paralegal, routine coordination, basic tech support - Low risk (<30%): Senior strategic roles, relationship-intensive roles, specialised expertise, management, physical work
By geography: - Highest impact: English-speaking tech economies (US, UK, Singapore, Australia, Canada) - Medium impact: Western Europe (union protections reduce speed) - Lower impact: Developing economies (lower AI infrastructure, language barriers, cost advantages make automation less economically viable)
By industry: - Highest: Financial services, customer service, technology, retail - Medium: Legal, manufacturing, healthcare admin - Lower: Healthcare clinical, government, skilled trades, management
If you are in an automation-pressure sector, the data shows an 18-36 month action window. Get your personalised assessment. Free score →
Get Your Free Risk Score