Predictions about AI replacing jobs are mostly vibes. Every columnist has a list. Most lists are wrong because they're not grounded in actual data about which tasks AI is actually doing, how much of which jobs are those tasks, and how exposed the economics are to automation.
The good news: by 2026 we have real data. Anthropic's Economic Index, McKinsey's Generative AI workforce reports, Brookings' occupational exposure analysis, and the BLS's latest Occupational Outlook all converge on roughly the same picture. Here's the honest ranking, with receipts.
The framework that actually works
Three factors determine how exposed a job is to AI in 2026:
- Task automatability. How much of the job is tasks current AI can do reliably? Text generation, summarization, structured data work = high. Physical work, judgment under uncertainty, relationship work = low.
- Output verifiability. Can AI output be checked quickly and cheaply? If yes, automation scales. If verification is expensive (like in medicine or law), humans stay in the loop longer.
- Economic structure. Is the job priced at a level where automation pays off? A $30/hour analyst is worth automating; a $300/hour judgment worker often isn't, because mistakes are too expensive.
Using these three lenses on Anthropic's Claude usage data (which occupations use Claude most heavily for which tasks), here's the ranking.
Tier 1: Highest exposure — significant automation happening now
Copywriters, content writers, technical writers. Anthropic's Economic Index shows content writing is among the most automated task categories on Claude. BLS data confirms the hit: writer/author roles grew 0% in 2023-2024 after years of steady growth, and average freelance rates have dropped 15-25%. The job isn't gone, but volume per head has collapsed.
Translators. Machine translation has been improving for a decade; GPT-4-class models pushed it over the quality threshold for most business use cases. BLS projects 4% growth 2022-2032 — but real demand is dropping, with growth concentrated in specialized legal, medical, and literary translation.
Customer support (tier 1). Tools like Intercom Fin and Zendesk AI handle 60-80% of inbound tickets at well-configured SaaS companies. The human work is shifting to complex escalations and agent oversight, not volume handling.
Junior paralegals and legal researchers. Harvey AI, CoCounsel, and Lexis+ AI are doing document review, research memo drafting, and contract analysis at partner-approved quality. A16z has written publicly about the legal tech category's margins and the headcount compression happening at mid-size firms.
Data entry and basic bookkeeping. This category was already shrinking before AI; generative tools accelerated it. BLS projects a 4% decline in bookkeeping roles by 2032.
Tier 2: High exposure — role reshaping, not replacement
Junior software engineers (CRUD work). Greenfield boilerplate code is now AI's sweet spot. Entry-level postings are down meaningfully in 2025. But senior engineering roles are up, and AI engineering is the fastest-growing job title in the US (+143% YoY). The category is bifurcating, not collapsing.
Junior analysts (BI, marketing analytics, FP&A). Dashboard-building and ad-hoc reporting are heavily automated. Senior analysts with statistical depth, experiment design skills, and business judgment are more valuable than ever.
Recruiters and sourcers. LinkedIn Recruiter AI, Gem, and HireEZ have automated large chunks of sourcing and screening work. Recruiter headcount is flat-to-declining; recruiting operations and exec recruiter roles are growing.
Marketing coordinators and execution managers. Campaign setup, copy variation, reporting, email personalization — all automatable. Strategic marketing roles (positioning, pipeline ownership, brand) growing in value. See the marketing manager AI analysis for the full split.
Tier 3: Moderate exposure — productivity gains, not replacement
Product managers. Drafting and synthesis work is automated; judgment, taste, and stakeholder work aren't. PMs who lean into the second category are fine. See the full PM analysis.
UX designers. Production UI work compressed; research, judgment, and cross-functional design hardening. Entry-level UI designer roles shrinking; senior product designer roles growing. Full UX analysis here.
Financial analysts (corp dev, IB, FP&A). Modeling and comps automated; judgment, sector expertise, and management interaction remain. Deep dive on financial analysts.
Mid-level accountants. Auditing, reconciliation, and tax prep increasingly AI-assisted. CPA judgment work and client-facing advisory roles growing.
Tier 4: Low exposure — mostly safe for now
Senior engineering, design, and PM leadership. Judgment-heavy, relationship-heavy, context-heavy work. Pay going up.
Sales (complex B2B). Relationship work, negotiation, and deal judgment. AI augments (call summaries, CRM automation) but doesn't replace.
Therapists, physicians, nurses. Regulatory, relational, and physical components. Some diagnostic AI assistance, but not replacement.
Skilled trades (electricians, plumbers, HVAC, welders). Physical embodiment required. BLS projects 6-8% growth across these categories through 2032.
Executives and senior strategists. Judgment, accountability, and relationship work.
What the aggregate data says
McKinsey's 2025 report estimates generative AI could automate 30% of tasks across US occupations by 2030 — but "task automation" is not "job elimination." Their same report projects net employment growth across most occupations, with the composition of work shifting. Brookings' analysis finds that high-exposure doesn't mean job loss; it means role reshaping. The BLS projects total US employment growth of 4.7% from 2022-2032, with faster growth in AI-adjacent technical roles.
The honest pattern: AI is a productivity multiplier, not an employment destroyer — but the distribution of gains is wildly uneven. Senior and judgment-heavy roles benefit. Junior and execution-heavy roles get compressed. The middle is disappearing fastest.
Four things to do if you're on this list
-
Figure out which side of the split you're on. Is your week mostly the automatable half or the durable half of your role? Audit honestly.
-
Move up the value chain, fast. Push for judgment work, strategy work, cross-functional work. Delegate the automatable work to AI tools yourself before the company does.
-
Invest in the skills that compound. Evaluation thinking, domain expertise, communication, stakeholder management. These don't depreciate when a new model ships.
-
Keep applying. The market is moving fast. If you haven't interviewed in 18+ months, you don't know what your market rate actually is. The perfect resume guide for 2026 and a tool like Rolevanta can get you back in the job market in a weekend.
The honest takeaway
"AI will replace my job" is usually wrong. "AI will eliminate the 40% of my job I was least excited to do, and the bar on the remaining 60% will rise sharply" is usually right. That's not a disaster — it's the single biggest career opportunity in a generation for people who adapt deliberately. The people who treat it as a disaster are the ones who aren't going to.
