Overview
Short version: AI is changing a lot of jobs, but not deleting most of them overnight. The core through‑line is a two‑speed reality — technical capability sprints ahead while organisational adoption moves slower and stepwise. The biggest visible hit is at the entry level, where graduate and junior roles have thinned out, while people who actually use AI at work are getting paid more. Inside companies, adoption is uneven — plenty of pilots, fewer full rewires — which is why impacts look lumpy rather than apocalyptic.
The Two‑Speed Reality (Core Pattern)
Technology capability
Frontier models keep leaping — multimodal reasoning, tool‑use, agents. Macro potential is wide: ~60% of jobs in advanced economies are exposed in some way.
IMF exposure anchor. Source
Organisational adoption
Reality is lumpy: lots of pilots, fewer end‑to‑end rewires. Hiring signals show the drag — high‑exposure job ads fell harder; entry‑level postings shrank; wage premiums appear where firms actually embed AI.
McKinsey (high‑exposure ads), Adzuna (entry‑level squeeze), PwC (skills premium).
Hiring split
Early‑career workers (22—25) in the most AI‑exposed occupations saw a ~13% relative employment decline since late 2022; experienced workers in the same roles held steady or grew.
Stanford Digital Economy Lab (Aug 26, 2025). Source
Wage/productivity split
Comp tight, employment adjusts first. Where AI is embedded, productivity and wages rise for AI‑literate staff — the skills premium becomes the real differentiator.
PwC AI Jobs Barometer (2025).
Original Forecast Snapshot (2023/24)
We projected pacey displacement in low‑resistance sectors and slower shifts in regulated ones. The page displayed tidy daily-loss counters and sector percentages. Useful framing, but some numbers came from weaker secondary aggregators and haven't aged well.
2025 Evidence
UK entry‑level postings are down roughly a third since late 2022. Roles with high gen‑AI exposure saw much larger drops in job ads than low‑exposure roles. Workers with AI skills are earning a substantial wage premium, and productivity gains are concentrated where AI is actually embedded. The IMF's anchor remains that ~60% of jobs in advanced economies are exposed — exposure means task change, not automatic replacement.
Sources further below: Adzuna (press), McKinsey UK (Jul 2025), PwC (2025), IMF (2024).
UK Snapshot
Entry level
Graduate, junior, and apprenticeship postings are down ~32% versus the pre‑ChatGPT baseline. The pipeline into white‑collar work is tighter, not gone.
Adzuna via UK press (Jun 30, 2025).
Exposure gradient
Vacancies in high‑exposure roles (analysts, programmers, marketing managers) fell ~38% vs ~21% in low‑exposure roles since 2022.
McKinsey UK (Jul 2025).
Skills premium
AI‑fluent workers command sizable wage premiums (often ~40—56%). Productivity effects show up where firms move beyond pilots.
PwC AI Jobs Barometer (2025).
Macro exposure
~60% of jobs in advanced economies are exposed to AI in some way; roughly half of those risk negative pressure without adaptation.
IMF (2024).
Scenarios to 2030
No single-number predictions here — just plausible bands and what would push reality around.
Base case
Through 2027, entry‑level demand in AI‑exposed office functions sits ~15—30% below 2022. Firms rebuild junior tracks with "apprentice + AI." Net headcount in those functions trends ~5—10% lower at large firms, mainly via slower hiring.
Upside
Workflows actually get rebuilt end‑to‑end. Junior roles return slimmer but richer. Productivity gains show up in revenue‑per‑employee and pay. The AI skills premium persists.
Drag
Adoption keeps stalling at pilot stage. Managers lean on automation to cover gaps without reskilling. The entry bottleneck lasts into 2028—2029. Routine mid‑tier roles hollow out; experienced AI‑literate staff remain in demand.
Confirmation Log
This section tracks new evidence that confirms or adjusts our earlier thinking.
Stanford Digital Economy Lab (Aug 26, 2025): ~13% relative employment decline for ages 22—25 in the most AI‑exposed occupations; adjustments show up more in employment than wages. Source: Brynjolfsson, Chandar & Chen (2025)
PwC AI Jobs Barometer (2025): Large and growing wage premium for AI skills; faster productivity growth in AI‑exposed sectors. Report
McKinsey UK (Jul 2025): High‑exposure job‑ads down ~38% since 2022 vs ~21% for low‑exposure roles. Blog
Adzuna via UK press (Jun 30, 2025): Entry‑level postings ~—32% vs pre‑ChatGPT baseline. Guardian · Independent
IMF (Jan 2024): ~60% of jobs in advanced economies exposed to AI; about half of those may face negative effects without adaptation. Blog · SDN 2024/001
Implications
Individuals: get fluent with the tools in your lane and document real output gains — that's what employers are paying for.
Teams: don't bolt AI on; redesign workflows. Pair juniors with AI‑augmented seniors and write the playbooks down.
Policy: focus on the entry bottleneck — apprenticeships, rapid upskilling, and public‑sector exemplars in high‑friction domains like health and education.
Sources & Archive
Primary 2024—2025 sources: Stanford DEL (2025) · PwC AI Jobs Barometer (2025) · McKinsey UK (Jul 2025) · Guardian/Adzuna (Jun 30, 2025) · IMF (Jan 2024).
Archived research packs (2023—2024): Claude · Grok · GPT‑4.5 · Original projections page.
This flagship page replaces earlier versions. Older pages remain online for transparency and historical comparison.