How vulnerable is your career to automation?

Most economists now believe AI will transform jobs rather than wholesale replace them. Even high-risk roles typically have some protected tasks. Radiologists may lose image-reading to AI but gain time for complex case consultation. The bigger risk is that companies restructure headcount rather than retrain, particularly in administrative roles.

The most protected skills are: physical dexterity in unpredictable environments, emotional intelligence and empathy, creative and novel problem-solving, leadership and persuasion, ethical judgment in complex situations, and interdisciplinary knowledge combination. Upskilling in AI tool use can also shift workers from "at risk" to "AI-augmented."

The WEF Future of Jobs Report 2025 estimates that 85 million jobs may be displaced by automation by 2026 but 97 million new roles will emerge. The transition period is the hardest, particularly for workers in mid-career. The McKinsey Global Institute projects 14% of workers globally may need to switch occupational categories by 2030.

The Frey and Osborne (2017) model remains the most widely cited academic framework for occupation-level automation risk but has known limitations. The original study assessed technical feasibility, not economic or political likelihood. A job scoring 0.95 does not mean 95% of people in that role will lose their jobs: it means 95% of the occupation's task bundle could theoretically be performed by machines. In practice, adoption depends on cost, regulation, social acceptance, and employer inertia. The OECD's 2023 task-based approach addresses some limitations by looking at individual tasks rather than whole occupations. Source: Frey and Osborne 2017, Oxford Martin School.

The occupations scoring lowest on the automation index share three characteristics: they require complex social interaction, creative problem-solving in unpredictable environments, or fine motor skills in unstructured physical spaces. Recreational therapists, dentists, athletic trainers, clergy, and registered nurses are among the least automatable. These roles involve empathy, real-time human judgment, and physical dexterity in variable conditions. Notably, some high-paying knowledge work scores higher than expected: accountants, paralegals, and insurance underwriters are technically more automatable than fast food cooks because their core tasks are data processing and rule application. Source: Frey and Osborne 2017.

As of 2026, AI has not eliminated entire occupations at scale, but it has significantly reduced headcount in specific task-heavy roles. The clearest examples are in customer service (chatbots handling 60-80% of tier-1 queries), data entry, and basic content generation. Translation services, basic legal document review, and routine financial analysis have all seen measurable workforce compression. The McKinsey Global Institute estimates that generative AI could automate tasks equivalent to 11.8 million US jobs by 2030, but projects that most displaced workers will transition to adjacent roles rather than face unemployment. Source: McKinsey Global Institute 2023.

A high automation score is not a redundancy notice. It is a signal to start strategic upskilling now, while the transition is gradual. First, identify which specific tasks within your role are most exposed. Second, focus on developing the tasks that remain human: client relationships, creative judgment, and complex problem-solving. Third, build AI literacy so you can use AI tools effectively in your current role, which makes you more valuable rather than more replaceable. Fourth, consider adjacent roles that use your domain expertise but combine it with less-automatable skill sets. Career transitions are easier when you start them before they become urgent. Source: WEF Future of Jobs Report 2025.

The original Frey and Osborne model predates large language models and generative AI. It significantly underestimates automation risk for roles involving writing, analysis, coding, and creative tasks. McKinsey's 2023 update specifically addresses this gap: their revised estimates show that generative AI has increased the automation potential of language-based tasks by 15-25 percentage points. Communication and persuasion tasks jumped from approximately 15% automatable to 38%. Occupations like copywriter, paralegal, and junior analyst are more exposed than the headline score suggests, while hands-on trades remain largely unaffected by the generative AI wave. Source: McKinsey 2023.

Every major technological transition in history has created more jobs than it destroyed, though with significant transition pain. The World Economic Forum's Future of Jobs Report 2025 projects that AI will create 97 million new roles globally by 2030 while displacing 85 million, for a net gain of 12 million. The new roles cluster around AI development, data analysis, digital transformation, and human-AI collaboration. However, the skills required for new roles rarely match displaced workers' skills, creating a mismatch that requires active retraining. Young, educated, urban workers adapt faster, while older workers in routine roles face the highest displacement risk with the fewest retraining options. Source: WEF Future of Jobs Report 2025.

Software engineering is a nuanced case. Frey and Osborne assign a moderate automation probability (approximately 0.48) to software developers, which is counterintuitive given that AI excels at code generation. The nuance is in the task decomposition: routine code generation, boilerplate writing, test generation, and documentation have high automation potential, and tools like GitHub Copilot already handle significant portions of these tasks. The higher-value tasks in software engineering — system architecture, requirements analysis, novel algorithm design, debugging complex distributed systems, and translating ambiguous business needs into technical specifications — remain difficult to automate because they require combining technical knowledge with contextual judgment, stakeholder communication, and the ability to navigate organisational complexity. McKinsey's 2023 analysis found that while 25-30% of software engineering time could be automated, this was likely to increase demand for software engineers by raising productivity and enabling more ambitious software projects, rather than to reduce employment. The consensus view among researchers and practitioners is that AI will substantially change what software engineers do rather than eliminate the role.

Estimates vary significantly by methodology and assumptions. The World Economic Forum's Future of Jobs Report 2025 estimated that AI and automation would displace approximately 85 million jobs globally by 2025 and create approximately 97 million new roles — a net positive, but with significant transition costs for workers in displaced roles. Goldman Sachs (2023) estimated that generative AI could eventually automate tasks equivalent to approximately 300 million full-time jobs globally, though with the caveat that historical precedent suggests most displacement leads to task reallocation rather than unemployment. McKinsey's 2023 scenario analysis projected that 30% of hours worked could be automated by 2030 in the US, but that employment would likely remain stable due to compensating demand effects. The honest answer is that no one knows with confidence how many jobs will be eliminated by 2030: the pace of AI capability development, regulatory responses, and economic adaptation are all highly uncertain. What the data does support is that the jobs most at risk are those with high proportions of routine cognitive tasks, that transition costs will be unevenly distributed, and that retraining and upskilling will be economically critical for displaced workers.