The 2030 AI Job Map: Which Roles Disappear, Which Survive

Every forecast about AI and 2030 falls into one of two shapes. The first is a headline: 300 million jobs will disappear. The second is a reassurance: new jobs will more than replace what is lost. Both are true at the aggregate level and useless at the personal level. The question that actually matters to you is narrower and more answerable: by 2030, which part of your week will still exist, who will be doing it, and what will it be called?

This post pulls together what the major research houses — the World Economic Forum, McKinsey, Goldman Sachs, the OECD, the IMF, the U.S. Bureau of Labor Statistics, and Anthropic — have actually said about 2030. It maps roles into four categories: disappearing, compressing, reshaping, and expanding. And it gives you a way to locate yourself on that map before the map locates you.

If you want the personal answer rather than the macro one, run our free AI career risk assessment and get your number in about 90 seconds. If you want the macro picture with sources, keep reading.

The pull-quote version: The 2030 job map is not about which jobs die. It is about which jobs fragment — which tasks leave, which stay, and whether what is left justifies a full-time role at the salary you want.

What the 2030 Numbers Actually Say

Before predicting which roles disappear, it is worth being precise about what the forecasts claim — and what they do not.

Goldman Sachs (March 2023, "The Potentially Large Effects of Artificial Intelligence on Economic Growth") estimated that generative AI could expose the equivalent of 300 million full-time jobs globally to automation. The widely quoted number is 300 million. The less widely quoted caveat from the same report is that roughly two-thirds of current jobs are exposed to some degree of AI automation, but only a portion of tasks within those jobs — Goldman's central case was 25–50% of task content — are actually automatable. "Exposed" is not the same as "lost."

McKinsey Global Institute (July 2023, "Generative AI and the Future of Work in America") put a cleaner 2030 number on it: up to 30% of the hours worked across the U.S. economy could be automated by 2030, with generative AI pulling the timeline forward by roughly a decade compared to their 2017 projection. McKinsey's working estimate was that 12 million occupational transitions may be needed in the U.S. alone by 2030 — meaning 12 million workers changing occupations, not 12 million unemployed.

The World Economic Forum's Future of Jobs Report 2023 — based on surveys of employers covering 673 million workers globally — projected that by 2027, structural labor market churn would displace 83 million jobs and create 69 million, for a net decline of 14 million (about 2% of employment). The 2025 update extended the horizon: 170 million new jobs created, 92 million displaced, for a net gain of 78 million by 2030. Employers expect 44% of workers' core skills to be disrupted by 2028.

The OECD (2023 Employment Outlook, "Artificial Intelligence and the Labour Market") found that 27% of jobs across OECD countries are in occupations at highest risk of automation, meaning occupations where more than 25 of 100 skills and abilities are highly automatable. Another 32% face significant change without full replacement.

The IMF (January 2024, "Gen-AI: Artificial Intelligence and the Future of Work") estimated that 40% of jobs globally are exposed to AI, rising to 60% in advanced economies — with roughly half of that exposure showing up as complementarity (AI augments the worker) and half as substitution (AI displaces the worker).

The U.S. Bureau of Labor Statistics' Employment Projections 2023–2033 (released September 2024) projected that total U.S. employment would still grow by 6.7 million jobs by 2033, but with significant occupational reshuffling — the fastest-growing roles are concentrated in healthcare, renewable energy, computer science, and care work, while projected declines cluster in clerical, routine administrative, and some retail roles.

Anthropic's Economic Index (February 2025 and subsequent updates) measured actual Claude usage patterns and found that AI is now used in at least 25% of tasks in about 36% of occupations, and in more than half of tasks in roughly 4% of occupations. This is a real-behavior counterpart to the theoretical exposure studies.

The convergence across these seven sources is narrower than the headlines suggest. The reasonable central estimate for 2030 is this: roughly 25–30% of current work tasks automatable, 10–15% of current occupations facing genuine role-level replacement risk, and 40–60% of workers facing significant task reshuffling without losing their jobs outright. The majority of the workforce will not be replaced. A meaningful minority will.

Pull-quote: 25–30% of tasks automatable. 10–15% of occupations at genuine replacement risk. 40–60% of workers facing task-level reshuffling. Those are the 2030 numbers after you strip the marketing from the research.

The Four Categories on the 2030 Map

Roles land in one of four categories by 2030 depending on two variables: how much of the work is automatable, and how fast the displaced workers can be absorbed elsewhere. We use those two axes to sort the map.

Category 1 — Disappearing: roles that shrink by 30%+ in headcount by 2030

These are occupations where more than half the week is automatable and the remaining tasks do not require the protected attributes — physical presence, regulatory licensure, trust built over time, accountability under ambiguity.

The WEF 2023 report's explicit "fastest declining" list for 2023–2027 was concentrated here and remains the best sourced forecast for the category: bank tellers, data entry clerks, administrative and executive secretaries, accounting/bookkeeping/payroll clerks, material-recording and stock-keeping clerks, cashiers and ticket clerks, and postal service clerks. The BLS 2023–2033 projections corroborate most of these — bank tellers -15%, word processors and typists -37%, data entry keyers -11%, executive secretaries and administrative assistants -21%, file clerks -11%.

Beyond clerical, two categories move into Category 1 by 2028–2030 based on current trajectory:

• Tier-1 customer service where interactions are bounded and scriptable. Klarna's 2024 disclosure that its AI agent handled customer-contact volume equivalent to 700 human agents has been followed by similar deployments at Shopify, Zendesk's "Answer Bot," Salesforce Service Cloud, and almost every large contact center. Our deeper analysis is in Will AI Replace Customer Service Representatives? — the durable portion is the escalation, de-escalation, and exception-handling work, not the front door.
• Junior-only variants of analytical and review work — roles where the whole job description is "extract information from documents, put it into a template, hand it up for review." Legal document review at the junior-associate tier, junior financial analyst model population, junior data analyst SQL generation, first-pass recruiting screeners. The senior versions of these roles are not in Category 1. The junior-only variants are. See Will AI Replace Lawyers? and Will AI Replace Financial Analysts? for the task-level breakdowns.

The disappearing category is real but smaller than the headline coverage suggests. The BLS projects roughly 1.6 million net job losses across clerical and routine-administrative occupations by 2033. That is meaningful, but it is roughly 1% of U.S. employment — not the apocalypse framing implies.

Category 2 — Compressing: roles that survive with fewer people per unit of output

Compressing roles do not disappear. They concentrate. One senior professional with AI leverage does the work three juniors used to do, and the middle of the career ladder thins out. Headcount falls 10–25%, but the job title persists.

Landed firmly in Category 2 by 2030:

• Software engineering — Microsoft, Google, and Meta have all disclosed that 25–30%+ of production code is now AI-generated in their environments. GitHub's Copilot is used by 1.8 million paid developers as of 2025. The role does not vanish; the junior-heavy leverage model does. See Will AI Replace Software Developers?.
• Marketing and content production — first-draft copy, campaign variants, social posts, and image generation are now produced in seconds. The surviving work is strategy, positioning, and taste. See Will AI Replace Marketing Managers? and Will AI Replace Graphic Designers?.
• Accounting, bookkeeping, financial analysis — the routine 60–75% of entry-level seats is automatable; advisory, audit defense, and judgment on complex treatment are not. See Will AI Replace Accountants? and Will AI Replace Data Analysts?.
• HR operations, recruiting coordination, project management — scheduling, screening, status reporting, and reporting-on-reporting compress sharply. Strategy, stakeholder management, and execution under ambiguity stay. See Will AI Replace HR Managers? and Will AI Replace Project Managers?.

Category 2 is where most readers of this post actually live. The 2030 risk is not "my job disappears." It is "my role survives, but the team shrinks, promotion to the next level is harder because the junior pipeline is thinner, and the required skill mix changes faster than I can keep up with."

Category 3 — Reshaping: roles where the work changes but headcount holds

These roles are protected by something AI does not replicate cheaply: physical presence, licensure, trust, or accountability in regulated settings. The tasks inside the job change materially by 2030, but the headcount does not collapse.

Strongly Category 3 on the current trajectory:

• Clinical healthcare — nurses, physicians, physical therapists, dentists. AI takes documentation, coding, triage support, and first-draft notes. It does not take bedside presence, hands-on procedures, or the clinical-legal accountability tied to the license. See Will AI Replace Nurses?. BLS projects registered nursing to grow 6% by 2033 and home health and personal care aides to be among the fastest-growing occupations overall.
• Teaching and training — K-12, special education, and hands-on vocational training are protected by the same compact between student, teacher, school, and parent. AI reshapes planning, grading, and differentiation — not the classroom itself. See Will AI Replace Teachers?.
• Skilled trades — electricians, plumbers, HVAC, carpentry. Aggregate hiring in the trades is constrained by people entering the field, not by labor replacement. BLS projects electricians +11%, plumbers +6%, wind turbine service technicians +60% through 2033.
• Real estate agents, financial advisors, senior B2B sales — trust-mediated roles where the purchase involves material risk to the buyer. See Will AI Replace Real Estate Agents?.

Category 4 — Expanding: roles that grow because of AI, not despite it

The fourth category is the smallest in absolute headcount but the most important for anyone choosing a pivot.

WEF 2023 and 2025 reports both identify the same fastest-growing roles: AI and machine learning specialists (+40% over five years in the 2025 update), data analysts and scientists, information-security analysts, renewable-energy engineers, solar installation and wind turbine technicians, electric-vehicle specialists, sustainability specialists, and — consistently — care workers and mental-health professionals.

BLS's 2023–2033 projections align: information security analysts +33%, data scientists +36%, statisticians +11%, software developers +17%, medical and health services managers +29%, nurse practitioners +46%, mental health counselors +19%, physical therapists +14%.

Two less-obvious pockets worth knowing:

1. AI-adjacent operating roles inside non-tech companies. Every mid-size company that adopts AI needs people who can integrate tools into real workflows, train teams, write internal prompts, monitor outputs, and own the vendor relationship. This role often does not have a standard title yet. It tends to be absorbed into existing operations, product, or IT functions. If you already sit in one of those functions, the upgrade path is obvious.
2. Care and trust-mediated work. The BLS growth list is dominated by it. Aging demographics in the U.S., Europe, and East Asia mean the demand for clinical and social-support work grows regardless of AI progress — and AI cannot replicate presence.

How to Locate Yourself on the 2030 Map

The map only matters if you can place yourself on it. A two-step self-audit gets you most of the way there.

Step 1 — Break your typical week into tasks, not hours. List the 15–25 distinct things you actually do in a normal week. Not "work on the Q3 plan." More like: "pull the sales dashboard, write the three-bullet summary, email it to the VP." Each task should be specific enough that you could, in principle, time it.

Step 2 — Sort each task into one of three buckets.

• Automatable by 2030. Digital input, digital output, acceptable error tolerance, a rulebook or training set exists. If you can imagine handing this to a competent intern with a good LLM and a clear brief, it is in this bucket.
• Augmentable. AI helps you do it 2–5× faster or better, but you still own the output and make judgment calls along the way.
• Protected. Physical presence, license, trust built over time, or accountability under ambiguity are required to deliver it.

If more than 60% of your tasks fall in the first bucket, you are in Category 1 or 2 territory — the role is at genuine risk, and the question is whether to compete on the top tier of it, pivot toward protected work, or pivot into an AI-native version of your function.

If 40–60% are augmentable and the rest are protected, you are solidly Category 2 or 3. The 2030 move is to become the person who orchestrates the AI for your function, not the person whose tasks get absorbed by it.

If most of your week is already protected, you are Category 3 or 4. The 2030 move is to adopt AI for the automatable edges so you do not pay the productivity tax.

This is the core methodology behind the paid riskquiz.me report — the free quiz gives you the score, the paid report gives you the task-level plan. The full methodology walks through the data sources and scoring if you want the technical version.

Three Things to Do Before 2030

Specific enough to act on this week, broad enough to apply regardless of role.

1. Move one layer up the value chain in your current function. If you write SQL, become the person who frames the question. If you write copy, become the person who owns the campaign. If you review contracts, become the person who negotiates. AI is cheap at the output layer and expensive at the framing and judgment layer — by 2030, the pay gap between those layers doubles.

2. Adopt the two or three AI tools that are already table-stakes in your function. Not every tool in the market — the two or three your senior colleagues and competitors already use. For most office roles in 2026 that means a top-tier general assistant (ChatGPT, Claude, or Gemini), a domain-specific tool (Harvey for law, CoCounsel or a comparable tool for legal/finance research, GitHub Copilot or Cursor for engineering, Jasper or equivalent for marketing), and whatever your employer has standardized on. If you cannot name the three, you are not yet defensibly competitive for a 2030 seat.

3. Start one protected thread. Something in your portfolio of work that is defensible on presence, trust, accountability, or licensure. For a knowledge worker this often looks like building a repeat-client advisory relationship, a public reputation in a specific domain, or a credential (CFA, PE, bar, clinical license) that gates the work. The protected thread is what lets you survive the compression in the automatable part of your job.

If you are unsure which category you fall into, the AI career risk assessment will place you on the map and tell you which of the three moves is highest-leverage given your specific answers.

FAQ

Q: What jobs will AI replace by 2030? A: The best-sourced forecasts — WEF 2023/2025, McKinsey 2023, BLS 2023–2033 — converge on clerical and routine-administrative work: data entry, bookkeeping and payroll clerks, bank tellers, executive and administrative assistants, postal clerks, file clerks, and cashiers. Add to this the junior-only variants of document review, basic analytical reporting, first-pass recruiting, and tier-1 customer service. Roughly 10–15% of current occupations face genuine role-level replacement risk by 2030. Most other roles compress or reshape rather than disappear.

Q: Will AI create more jobs than it destroys by 2030? A: The WEF's 2025 Future of Jobs Report projects 170 million new jobs created and 92 million displaced globally by 2030 — a net gain of 78 million. The IMF, OECD, and BLS broadly agree that net employment grows through 2030. The catch: the new jobs are concentrated in AI/data, healthcare, care work, renewable energy, and skilled trades, while the displaced jobs are concentrated in clerical and routine administrative work. Net growth at the aggregate level coexists with painful transitions at the individual level, especially for workers without reskilling access.

Q: Which jobs are safest from AI by 2030? A: Three categories survive 2030 well. First, roles protected by physical presence — skilled trades, clinical healthcare, personal care, complex hands-on work. Second, roles protected by licensure and accountability — medicine, law at senior levels, engineering sign-off, financial advice under fiduciary duty. Third, roles protected by trust and relationships — senior sales, advisory work, therapy and counseling, and leadership where stakeholder management is the actual product. BLS's fastest-growing 2023–2033 list is dominated by these categories.

Q: Should I change careers before 2030 because of AI? A: For most people, no — the stronger move is to change how you do your current career rather than the career itself. The pattern in the compressing roles (most office work) is that the senior, AI-augmented version pays more than ever and the junior, pre-AI version pays less and hires less. If you can move up the value chain inside your function and adopt the two or three table-stakes AI tools, you keep the role and gain leverage. Full career pivots make sense when your role is in Category 1 (genuine replacement), when you lack the inputs to reach the senior tier of your current function, or when you have a clear path into an expanding category. The free assessment is designed to surface which of these applies to you.

The 2030 Move

The 2030 job map is not a prediction about whether AI will be powerful enough to replace you. It is a prediction about whether the market around you will decide that one senior human with AI can replace three junior humans without AI. In most functions, by 2030, it will. The question is which of the three you are.

Run the free AI career risk assessment if you want your personal answer in 90 seconds. Read the methodology if you want to see the data sources and scoring behind it. And if you want the full picture of how this plays out role by role, the hub piece on whether AI will take your job and the task-level breakdown are the two companions to this one. For the near-term view, the ten 2026 predictions extrapolate from what just happened in 2024–2025.

The jobs that disappear by 2030 are the ones where nobody bothered to do the audit in 2026. Yours does not have to be one of them.