What the current evidence actually says
The International Labour Organization’s 2025 global index found that one in four workers are in occupations with some generative-AI exposure, while only 3.3% of global employment falls into its highest exposure category. Its central conclusion is not that one in four jobs will vanish: because most jobs include tasks requiring human input, transformation is the more likely effect for most occupations. [1]
Why “exposed to AI” does not mean “replaced by AI”
Exposure measures ask whether an AI system could help with a task. That is very different from proving that an employer will remove a role, that a person can no longer build a career in that field, or that a tool can take responsibility for the full outcome. A software developer may use AI to draft code; an accountant may use it to classify a transaction; a teacher may use it to produce an activity. The job still includes decisions, context, collaboration, verification, and consequences.
That distinction matters because scary predictions can sound more certain than the data allows. The Yale Budget Lab’s October 2025 review found no discernible economy-wide disruption in the U.S. labour market attributable to AI in the first 33 months after ChatGPT’s release. That does not prove AI will have no effect. It does mean that claims of immediate, universal replacement go beyond the evidence. [4]
At the same time, it would be complacent to say nothing is changing. The OECD notes that current AI capabilities are closest to routine information processing, administrative work, and codifiable tasks, and furthest from work requiring contextual judgment, interpersonal understanding, complex decision-making, and responsibility. Actual outcomes still depend on adoption, regulation, workplace design, and social choices. [2]
Five signals that make a career more resilient
These are not a scoring system and they do not guarantee employment. They are practical lenses for comparing paths. The stronger the combination, the less likely a role is to be reduced to a simple prompt-and-output workflow.
Real-world complexity
The work changes with the body, building, machine, environment, or situation in front of you. An electrician diagnosing an old building and a nurse assessing a patient are not completing an identical task each time.
Human trust and care
The work relies on relationships, teaching, reassurance, persuasion, conflict management, or support. Tools can assist communication; they do not automatically earn trust.
High-stakes judgment
A poor decision can affect safety, health, money, rights, or a person’s future. In these settings, someone must examine the evidence, understand context, and explain the decision.
Regulation and accountability
The work requires a licensed, qualified, or otherwise accountable person to sign off, supervise, or take responsibility. AI can be part of the process without becoming the responsible party.
Adaptability
A resilient path gives you a foundation that transfers: learning new tools, moving into adjacent specialisms, combining technical and human skills, or taking on more complex problems over time. This may be the most important signal because every field will change.
Career paths worth investigating—by work pattern
Instead of chasing a list of “safe jobs,” investigate work patterns that fit you. The examples below are places to explore, not rankings or guarantees. Local training routes, wages, labour demand, and working conditions can differ dramatically.
Care and clinical responsibility
Roles such as nursing combine hands-on work, clinical observation, communication, ethics, and accountable decision-making. AI may support documentation and information retrieval, but it cannot independently deliver humane, safe care.
Explore the nursing degree guide →Skilled trades and physical systems
Electrical work, plumbing, HVAC, maintenance, and related trades take place in varied physical settings and involve safety, diagnosis, and regulated practice. Technology will change the systems being installed and maintained, not remove the need for competent people.
Explore the electrician apprenticeship guide →Teaching and human development
Teaching is more than generating learning material. It includes classroom leadership, safeguarding, motivating learners, adapting to confusion, and building a learning environment. AI can assist with resources; it cannot replace the whole relationship.
Explore the teaching degree guide →Design under real constraints
Architecture and built-environment work mix creativity with sites, regulations, materials, budgets, engineering, and long-term use. Generative images can speed up early exploration; they do not turn a concept into a safe, buildable project.
Explore the architecture degree guide →Careers changing fast can still be good choices
Some of the most AI-exposed paths are still worth considering. The right interpretation is not “avoid every job that uses a computer.” It is “understand which tasks are becoming easier, what new expectations are emerging, and whether you want to build the deeper capability around them.”
| Career pattern | What AI may accelerate | What the decision should focus on |
|---|---|---|
| Software and technical work | Code drafts, documentation, debugging support | Systems thinking, verification, product context, security, collaboration, and learning speed |
| Accounting and law | Research, document extraction, standard drafts, routine processing | Professional judgment, ethics, regulation, client needs, and accountability |
| Marketing, design, journalism | First drafts, variants, images, summaries, routine content | Original insight, research quality, audience understanding, editing, strategy, and trust |
| Care, teaching, and trades | Administration, preparation, records, and information support | Real-world context, relationships, safety, judgment, and responsible action |
The U.S. Bureau of Labor Statistics makes the same wider point in its 2024–34 projections: AI and IT adoption may boost demand in some occupations while dampening demand in others. It projects 33.5% growth for data scientists and 15.8% for software developers, while predicting declines in several office and administrative-support occupations where productivity gains may limit demand. Those are U.S.-specific projections, not a global career ranking, but they show why “technology job” versus “non-technology job” is too simple. [3]
How to choose without pretending you can predict 2035
You do not need to predict every new model, software feature, or labour-market shift to make a useful choice. You need a decision process that is robust when the future is uncertain. Start with the work you would be willing to practise: the problems you enjoy, the environment you can tolerate, the people you want to help, and the responsibilities you are willing to carry.
Look at the actual work. Read job descriptions, observe a workplace, talk to people in the field, and ask what a difficult day looks like—not only what the role is called.
Research the local route. Compare degree, apprenticeship, licensing, cost, work experience, and demand in the country or region where you expect to live.
Check the task mix. Ask what is routine and codifiable, what needs context and trust, and what a person remains accountable for.
Choose a foundation that can expand. Look for training that teaches principles, practical experience, communication, and ethical use of tools—not a narrow workflow that may disappear.
Test before you commit. A short course, an open day, a volunteer role, a work-experience placement, or a conversation with a practitioner can replace a lot of speculation with real evidence.
The goal is not to beat AI at being AI. It is to become someone who can use tools well, understand a real domain, work responsibly with people, and adapt as the tools change.
Frequently asked questions
References
- International Labour Organization, Generative AI and Jobs: A Refined Global Index of Occupational Exposure (2025).
- OECD, AI and Work: current research and policy evidence.
- U.S. Bureau of Labor Statistics, Artificial intelligence, information technology, and employment, 2024–34 (2026).
- The Budget Lab at Yale, Evaluating the Impact of AI on the Labor Market: Current State of Affairs (2025).
Aifyx provides general career information, not personal education, financial, or employment advice. AI exposure is not a prediction that a particular role will disappear. Employment statistics referenced on this page are clearly labelled by their source and geography.