AI Is Taking These Jobs First (Not the Ones You Think)

Let's talk about AI. AI is coming for your jobs. You've probably heard, read or seen that a million times by now. Remember when AI meant that bald white robot from iRobot? Simpler times. Can a robot write a symphony? Can a robot turn a canvas into a beautiful masterpiece? Can you? Will Smith and AI, huh? Now, the raging question amongst the many enraging debates today is does AI destroy jobs or does it create? Both sides are confident. They come armed with data to prove themselves. Both are also slightly wrong because they don't talk about any one specific job. Their data is about averages. Averages don't have careers. People do. Now, think about the last 24 hours at work. How did you spend them? What were the five things that you did that you could easily give to a model? And how many of those things needed the so-called human element? I think we need to change the question that we're asking. What we should be asking is which part of our jobs will AI take? Now, here's how professionals answer this question. And I think it makes sense. Researchers break each role into individual tasks and ask which of these tasks can AI already do really, really well? Across research done by Microsoft, PwC and India's own labor data, the pattern holds. The most exposed work is routine, templated, lookup-based. First drafts, basic coding, data entry, tier one support. In India, the squeeze is already measurable. Some surveys point to a 20-25% decline in entry-level tech hiring in parts of the sector. At Infosys, employees under 30 now make up just over half the workforce, down from 53% the year before. That is the lowest share in 15 years. Now, for a moment, that sounds like confirmation of your worst fear. Entry-level work is shrinking, the bottom rung is gone, but that's not quite what the data says. If you take a closer look, the takeaway isn't that AI is leading to less hiring. It's that AI is leading to hiring differently. Now, PwC's global data shows that AI-exposed entry-level roles are now seven times more likely to demand senior skills. Judgment, leadership, the ability to review and correct what the AI produces. Those seniorized entry-level roles have grown 35% since 2019. The latter didn't disappear. The first few rungs got harder and new ones appeared higher up. Now, there's three categories that keep on surfacing as genuinely safe when it comes to job security. Number one, roles involving physical work in unpredictable environments. Think construction, farming, food preparation. All of these sit at 0% high exposure in one major 2026 index. Now, AI can describe how to fix a pipe. It cannot go under one at two in the morning during monsoon season. Number two, roles involving high-stakes judgment with accountability. A doctor, a lawyer, a manager deciding brand identity. AI can draft a memo. It cannot own the decision. And lastly, relationship deep work. Negotiation, therapy, recruitment. Where the harder conversations still belong to people because that's where the actual job has always lived. Which raises the only question worth asking. How much of your day is task and how much of it is actual judgment? Now, don't panic, but don't stay in denial either. Spend less time on skills a model can already do and focus on the skills AI is not built to handle. See you in the next one.