Job market 2026: what's really driving AI layoffs and how to adapt

Executive overview

Major layoffs at Block and Atlassian are being attributed to AI, but insiders say over-hiring from the last tech boom is at least partly to blame. The real underlying shift is genuine: routine, rules-based office work is being automated, and hiring into AI-exposed roles is slowing — especially for newcomers.

The framework that matters is a two-layer job model: layer one (repeatable, templated tasks) is rapidly being taken over by AI; layer two (judgment, relationships, strategy) is growing in value.

The premium is rising on people who combine human skills, AI fluency, and real domain expertise.

The AI layoff narrative: what's real

• Block cut 40% of its workforce (4,000+ people); CEO Jack Dorsey explicitly cited AI and smaller, flatter teams
• Atlassian cut 1,600 roles, citing a changing skills mix due to AI
• An insider at the World Economic Forum noted that AI is also a "convenient excuse" to fix over-hiring from the 2021–22 boom
• US employers explicitly blamed AI for tens of thousands of job cuts in 2025 — sometimes real automation, sometimes rebranded cost-cutting

What the Anthropic research actually shows

• Anthropic's Labour Market Impact study measured observed AI use, not theoretical exposure
• For white-collar roles (software engineers, lawyers, finance, admin), AI can technically touch 90%+ of tasks
• In practice, workers currently use AI on only 30–40% of those tasks — but that share is growing every six months
• No major unemployment spike yet for AI-exposed roles; what's visible is a slowdown in hiring into those roles, especially for younger workers
• ~30% of workers have near-zero AI exposure today: cooks, mechanics, tradespeople, in-person services

The two-layer job model

• Every job has a task layer (layer one) and a judgment layer (layer two)
• Layer one: routine, rule-based work — writing templated emails, processing claims, scheduling, filling reports
• Layer two: context, relationships, intuition, experience — the work AI cannot reliably replicate
• If 80% of your day is layer one, your market value is quietly declining each year
• If 80% is layer two, AI is a force multiplier: it handles the routine, freeing you for higher-value work

The workforce reskilling picture (WEF data)

• Out of every 100 workers globally, ~50 need rapid reskilling by 2030
• Two-thirds of those could reskill within their current role
• One-third would need to move to a different role inside the same organisation
• About 11 out of 100 workers have no clear destination inside their current industry
• Declining roles: administrative assistants, parts of customer service
• Growing sectors: agriculture, education, healthcare — fields with global shortages that AI won't replace

What skills matter most

• Human skills — creativity, empathy, communication, leadership, social influence, self-management — are rising in value, not falling
• AI fluency — using tools like Claude, ChatGPT, Copilot for drafts, summaries, analysis, prompts, workflows
• Domain expertise — the context needed to judge whether AI's output is correct and turn it into real decisions
• Being "AI native" means defaulting to "what can I offload to AI?" regardless of industry or role
• The paradox: the more powerful AI becomes, the more valuable human skills become

Practical 90-day plan

1. Next 30 days: pick one AI tool and use it daily for your actual work — writing, summarising, planning, or analysis
2. Next 60 days: ship one small AI-powered improvement — an automated report, smarter template, or content workflow
3. Next 90 days: deliberately practise one human skill (communication, negotiation, leadership) in a real project with other people