Article 1: Why "Which careers are safe from AI?" is the wrong question

The question everyone is asking

Every parent making a career or education decision for their teenager now asks some version of the same thing: "Which careers will AI destroy? What should my child study to stay safe?"

It's a sensible question. It's also the wrong one to answer.

Not because the question is stupid—it's not. It's because the answer will almost certainly be wrong, and wrong in ways that will matter for your teenager's actual pathway.

Why the question fails

For a "safe" career question to work, you'd need three things:

First, you'd need to predict which tasks AI will handle. Not which jobs, which tasks. Because "accountant" still exists in 2035—but what accountants spend their time doing might change completely. AI may take over substantial parts of routine tax work or document review—or progress may concentrate somewhere else entirely. Predictions about this have a short shelf life and a worse accuracy rate than we usually admit.

Second, you'd need to know which of those tasks matter for the learning phase. A solicitor's work transformed by AI might still require apprentices to learn legal reasoning, even if the research work changes. Or it might not. We won't know until it happens—or shortly after it happens, when it's too late to change course.

Third, you'd need confidence that this prediction holds for your teenager's working life. If your teenager starts a three-year degree in 2026, they'll graduate in 2029 and work from roughly 2029–2070. Predicting with any confidence what "safe" means across a 40-year window is asking more of forecasting than forecasting can deliver.

We've seen this pattern before. In 2010, a parent might have confidently predicted which careers would thrive and which would struggle based on assumptions about smartphone technology and adoption. The recurring problem is that confident forecasts often underestimate second-order effects and unexpected changes. What actually matters is less "Will this job exist?" and more "How can my teenager build adaptability?"

What the evidence actually shows

DATA: What published research shows

Here's what we do know:

1. AI capability has expanded from theoretical to workplace-practical. In 2020, generative AI had relatively limited mainstream workplace use. Within a few years, tools capable of drafting, coding, analysis and document review had become widely available. In legal work specifically, research shows adoption moving quickly: LexisNexis reported that 41% of surveyed UK lawyers were already using generative AI in late 2024, with research, analysis and drafting among the primary use cases.

2. AI is affecting activities within jobs, not simply eliminating whole occupations. McKinsey estimates that activities accounting for up to around 30% of current working hours could be automated by 2030 in its modelling, accelerated by generative AI. But this effect is highly uneven: some routine, information-processing and drafting tasks are more exposed than activities involving judgment, responsibility and interaction with other people. Automation affects specific tasks, not whole professions.

What to ask instead

The more useful question isn't "Which careers are safe?" It's a cluster of better questions that focus on transferability, reversibility, and adaptability rather than prediction.