Approximate estimate. Employment-weighted average of Eloundou et al. (2023) β: the proportion of O*NET tasks where LLM + tools can reduce completion time by ≥50%. Based on 19,265 tasks across 923 occupations, labeled by human experts and GPT-4. Aggregated: Σ(β_occ × emp_occ) / Σ(emp_occ).
Eloundou et al. 2023 ↗§ I · The clock
Industry-Wide AI Progress: Where are we?
The career test tells you where you stand. This page is the board the game is played on — how fast AI is actually moving, what it has already taken, and what comes next.
Approximate estimate. Among tasks AI can perform, how often AI output quality ≥ human expert. 44 occupations measured directly via OpenAI GDPval (GPT-5.2 vs human, blind-judged); remaining ~730 occupations use sector or global average as fallback. Employment-weighted: Σ(P_occ × emp_occ) / Σ(emp_occ).
OpenAI GDPval (GPT-5.2) ↗World Economic Forum projects 92M jobs displaced by AI and automation by 2030.
World Economic Forum, 2025 ↗AI Replacement Progress+1.8%/mo
24.7%
↓ WE ARE HERE
AI Assist
You lead, AI executes
AI Augment
You lead, AI elevates you
AI Agent
You direct, AI works
AI Lead
AI leads, you support
Kill Threshold
Full replacement
§ II · The last mile
Your Data Is Training AI to Replace You
Skill is the last mile from general AI to your specific job. Your experience data is the last defense line.
Foundation Model
General capability, no specialization
Fine-tuning / Skills →Your Data & Experience
Industry knowledge, workflows, expertise
Specialization →Your Replacement
AI that does YOUR specific job
The more open your experience data is, the faster AI crosses the last mile to replace you.
§ III · The timeline
From Steam to AGI
250 years of accelerating change — and where you stand on it.
- 1769Steam Engine
Machines learned to move — replacing physical labor.
- 1879Electricity
Power distributed everywhere; factories ran 24/7.
- 2012Deep Learning
Neural networks reached human-level perception.
- 2022ChatGPT
100M users in 2 months — AI goes mainstream.
- 2025AI Agents
AI plans and executes multi-step work autonomously.
- 2026We Are HereNOW
AI already replacing 24.7% of work — and accelerating.
- 2030AGIPROJECTED
Human-level AI projected across all cognitive domains.
§ IV · First targets
AI's First Targets
AI doesn't hit everyone equally — it picks the soft targets first. If your work is rule-describable, repetitive, and document-heavy, you're on the priority list.
Customer Service
● High replacement
95%
Phone & online support
AI handles ~80% of standard queries by 2025.
Admin & Support
● High replacement
90%
Assistants, data entry
Document/data work is highly codifiable.
Finance & Accounting
● Mixed impact
65%
Junior analysts
Entry-level at risk; senior roles augmented.
Media & Content
● Mixed impact
50%
Copywriting, basic design
Low-end replaced; creative work enhanced.
Software Development
● Mixed impact
45%
Junior developers
Young devs −20%, overall +17.9% growth.
Manufacturing
● Collaboration
55%
Quality inspection, monitoring
Human-machine collaboration; no mass layoffs.
Education
● Augmentation
20%
K-12 teachers
BLS projects +5% growth (2024–2034).
Healthcare (Radiology)
● Augmentation
15%
Radiologists
BLS projects +5% growth; AI as assistant.
§ V · Not theory
Companies Are Already Cutting Jobs with AI
In the news it reads as a layoff number. In the company spreadsheet it reads as "taken over by AI and automation."
30,000cut
Automation & restructuring
Logistics4,500cut
Focus on AI & automation
Chemical775cut
Automating warehouses
Retail~15%cut
Shifting to AI-driven products
Tech276,000+cut
AI-driven restructuring (2024–25)
2024–25§ VI · The full picture
Displacement vs Creation
92M
jobs displaced by 2030
170M
new jobs created
+78M
net gain
Structural reshuffling, not total collapse. New jobs outnumber displaced ones — but the transition pain is real, and it lands first on the people on the list above.
World Economic Forum, Future of Jobs 2025
Where do you land?
Find your archetype
Take the test
Sixteen ways to stand against the clock.