Babith Bhoopalan spent 25 years at Microsoft, leading AI initiatives. When his 16-year-old daughter Thea set her sights on a career in finance, he did what most parents cannot: he went to the data.
The result was a 20-page document cross-referencing 17 sources, including the World Economic Forum, Goldman Sachs, McKinsey and Anthropic, scoring 35 careers across nine categories measuring vulnerability to automation. The findings are worth understanding whether you have a teenager at home or not.
The ladder is losing its bottom rungs
The most counterintuitive conclusion is also the most important. AI is not threatening careers from the top. It is removing the entry points first. Graduate intakes at the large accountancy firms have fallen sharply. UK entry-level job postings have dropped by close to a third since ChatGPT arrived. Junior contract review, financial modelling, document drafting: these are the apprenticeship stages through which a generation learned to become experienced. Many of them are going.
Where the resistance is strongest
Mental health therapists score 98% on Bhoopalan's AI-resistance framework, firefighters and paramedics 97%, surgeons and physicians 96%. Skilled tradespeople sit between 82% and 94%. The pattern across the highest-scoring roles is consistent: physical judgment in novel environments, trust built over time, and ethical decisions made under pressure. A blocked drain in a Victorian terrace is not a repeatable task. Neither is a therapeutic relationship developed over months with someone in genuine distress.
Law and finance divide sharply by seniority. Senior partners and fund managers who exercise judgment score well. Junior associates and graduate analysts doing process work score between 35% and 50%, and falling.
The wage premium is already visible
Workers with genuine AI fluency earn 56% more than those without it, across every field. That premium doubled in 12 months. AI-related job postings are growing three times faster than average. Bhoopalan's comparison is blunt: learning to use AI tools fluently in 2026 is what learning to type was in 1995.
His daughter Thea has switched from finance to international relations. Negotiating between human beings, building trust across cultural divides and making decisions under pressure scores in the 90th percentile for AI resistance. She worked that out herself.