What skills will be most in demand in the next five years?

The durable pattern: skills that involve judgement, integration, accountability or physical presence are appreciating; skills that are routine information processing are depreciating. Rather than a list of technologies — which age badly — that heuristic is worth more than any prediction. **Appreciating, with reasoning:** 1. **Working effectively with AI in a specific domain.** Not 'prompt engineering' as a job, but being the person in your field who uses these tools well and knows where they fail. This applies to law, medicine, teaching, accounting, design and engineering alike, and it's the broadest opportunity available. 2. **Evaluation and verification.** Deciding whether output is actually correct, systematically. As generation becomes cheap, judgement becomes the bottleneck. This is currently the scarcest skill in AI work and it generalises well. 3. **Systems thinking and integration.** Understanding how parts fit together, what breaks when one changes, and where the real constraint is. Hard to automate because it depends on context that isn't written down. 4. **Communication, especially written.** The ability to explain a technical thing to a non-technical decision-maker, to write a clear proposal, to run a meeting that produces a decision. Chronically undervalued by technical people and consistently what separates senior from mid-level. 5. **Data literacy.** Reading a chart critically, knowing what a metric hides, spotting a misleading comparison. Increasingly required outside analytical roles. 6. **Skilled trades and hands-on technical work.** Electricians, technicians, healthcare workers, maintenance. Physical dexterity in unpredictable environments remains far from automated, and demographics are tightening supply. 7. **Security.** Persistent shortage, rising stakes, and the attack surface grows with every new system. 8. **Managing and building trust with people.** Genuinely hard to automate and the demand doesn't fall. **Depreciating**: routine content production, basic translation, simple data entry and transformation, first-line support scripts, entry-level analysis that follows a template, and any work whose output is 'a document produced by following a known procedure'. **The meta-skill worth most**: learning quickly. The specific tools will change several times in five years. People who can pick up a new system in weeks stay employable through every shift; people whose value is one tool are exposed each time it changes. And the practical advice: depth in one domain plus breadth across adjacent ones beats shallow coverage of whatever is trending. Being genuinely good at something specific, and able to work across boundaries, has outperformed trend-chasing in every previous technology cycle.