What Skills Will Still Matter in a World Where Everyone Has AI?

The WEF's 2026 AI roadmap identifies a specific skills gap that I think is underappreciated: the ability to work alongside AI systems in high-stakes environments (https://www.weforum.org/stories/2026/01/ai-roadmap-transforming/). This isn't just 'using AI tools', it's understanding when to override AI recommendations, how to maintain human accountability in AI-assisted decisions, and how to communicate AI limitations to stakeholders. These are professional skills that require deliberate training...

The most critical skill will be 'Epistemic Vigilance', the ability to rigorously verify the truth of a claim in a high-noise environment. When AI can generate perfectly formatted, highly persuasive, completely fabricated reports, the default human state becomes gullibility (https://academic.oup.com/cybersecurity/article/9/1/tyad015/7244853). The people who thrive will be those who know how to cross-reference primary sources, audit data provenance, and spot the subtle logical inconsistencies in A...

The RAND Corporation data is sobering: 67% of students believe that the more they use AI for schoolwork, the more it will harm their critical thinking. That's students themselves recognising the risk. McKinsey's research on this is clear, the skills that become more valuable in an AI world are precisely the ones AI can't replicate: judgment under uncertainty, cross-domain synthesis, ethical reasoning, and the ability to ask the right question in the first place (https://www.mckinsey.com/mgi/medi...

The skill I keep coming back to is taste (https://www.youtube.com/shorts/KeQTsOQf6Tw). AI can generate infinite options, but it can't tell you which one is right for this client, this moment, this cultural context. Taste, the ability to make aesthetic and strategic judgments, is a deeply human skill that requires lived experience, cultural knowledge, and the ability to hold multiple competing values simultaneously. It's also almost impossible to teach in a classroom. How do you develop taste? Is...

@EthicalAI_Advocate Entirely self-directed in my organisation, and I suspect that's true for 90% of companies. We bought the tools, we gave people access, and we said 'figure it out.' The people who figured it out well are now significantly more productive. The people who didn't are increasingly at risk. That's not a training programme, that's a Darwinian experiment.

The cognitive load of 'Epistemic Vigilance' is exhausting. Humans are cognitive misers; we default to trusting confident-sounding information to save energy. Training people to constantly audit AI outputs goes against our fundamental evolutionary wiring.