Stop Measuring Skills. Start Measuring the Capacity to Learn Them.

Somewhere in your organisation, someone is maintaining a skills taxonomy. A large, earnest, extremely well-formatted document- possibly living in a platform that cost more than it should have- that maps every role to every competency, every competency to every proficiency level, and every proficiency level to a suite of learning interventions that were almost certainly out of date by the time anyone finished building them.

This person is working very hard on something that is, I'm sorry to say, largely a waste of time.

Not because they're doing it wrong, but because the underlying model is wrong. A skills taxonomy is, so often, a static answer to a dynamic problem. It takes a photograph of what capability looks like right now, at this particular moment, in this particular market- and then spends eighteen months turning that photograph into a framework, by which point the photograph no longer resembles reality. The half-life of a specific professional skill is shrinking faster than any taxonomy can track. What distinguished a capable data analyst in 2022 is a given by 2025. What is cutting-edge AI practice today will be table stakes in eighteen months.

You cannot map your way out of a moving target. And yet here we are, mapping.

Carol Dweck's research on growth mindset has been cited so many times that it risks losing its edge, so let me try to claw some of it back. The finding that matters is not the headline version- "believe you can improve and you will"- which is useful but incomplete. The deeper finding is about what people actually do when they hit difficulty. People who believe their capabilities are fixed treat challenge as a verdict: I'm either good at this or I'm not. People who believe their capabilities are developable treat challenge as information: I'm not good at this yet, and that's the starting point for something. The word "yet" is doing an enormous amount of work.

What Dweck's research actually points at- and what the organisations who cite it most enthusiastically tend to miss- is that the most valuable thing you can develop in a person is not a skill. It is the habit of development itself. The ability to enter an unfamiliar domain, reach functional competence quickly, transfer learning across contexts, and keep going when it's uncomfortable. This is what researchers call learning agility, and it is the meta-competency- the one that stays valuable regardless of which specific skills the market happens to need next.

And here's the really special thing about learning agility: it is measurable. Yes, you read that correctly. It shows up in how people approach new situations, how quickly they reach competence in something unfamiliar. Whether they move toward gaps or paper over them. Whether, when they genuinely don't know something, they say so and get curious, or perform certainty and hope nobody notices. That last one, in particular, is enormously informative. It tells you more about someone's long-term trajectory than most formal assessments will.

Robert Sternberg spent much of his career arguing that we have a serious blind spot in how we think about intelligence and capability, and that it is costing us. His triarchic theory proposed that human intelligence is not one thing but at least three: analytical intelligence (the kind that does well in exams and IQ tests), creative intelligence (the ability to handle genuine novelty, make unexpected connections, operate well in ambiguity), and practical intelligence (the judgment to apply knowledge in real contexts- knowing which tool to reach for, when to trust an output, how to navigate the gap between theory and the specific messy and stressful situation in front of you).

Organisations, he observed, consistently over-select for the first and systematically undervalue the other two. Partly because analytical intelligence is easier to test. Partly because creative and practical intelligence are harder to put in a taxonomy.

In a world where AI can handle a growing proportion of analytical tasks- where you can get a competent first-pass analysis produced in minutes- what remains distinctively, irreducibly human is the judgment that surrounds it. The decision about what question to ask. The evaluation of whether the answer is actually right. The ability to notice when a novel situation requires a different approach entirely. These are expressions of Sternberg's neglected categories, and they are precisely the things no model can replicate. If your talent strategy is still primarily optimising for analytical skill, you are selecting for the thing AI does best, and underinvesting in the things it can't touch.

There is a version of this argument that ends with "therefore AI is bad" and I want to actively refuse that conclusion, because it is both wrong and, frankly, boring. AI is genuinely transformative and the organisations that figure out how to use it well will have a real advantage. The question is not whether to use it. The question is what you are using it for.

John Dewey- writing over a century ago, which is either reassuring or alarming depending on your disposition- drew a distinction between experience and educative experience. Experience, he said, simply happens. Educative experience is structured to produce growth: it connects to what came before, stretches the person in the right direction, and leaves them more capable of meeting what comes next. Most experience at work is not educative in this sense. It is repetitive. It maintains existing capability rather than building new capability.

The opportunity that AI creates- if we take it seriously- is to use the efficiency it provides to make more of the experience at work genuinely educative. If AI handles the routine analytical work, what does that free people up to do? Ideally, the higher-order judgment work that develops the kind of capability that matters. The creative thinking. The ambiguous decisions. The conversations that require genuine reading of another person. The problems that don't have obvious solutions. If we are intentional about this, AI becomes a development accelerant. If we are not, it becomes a way of producing better outputs from people who are not developing- which is comfortable in the short term and quietly catastrophic over time.

In practice, the shift from a skills-based to a learning-agility-based talent strategy is less dramatic than it sounds. It does not mean abandoning domain knowledge or pretending that specific skills don't matter. A surgeon still needs to know anatomy. The question is whether anatomy is sufficient as the primary unit of measurement, and the answer is increasingly no.

It means changing what you pay attention to in hiring: not just what someone knows but how they have developed, how they've responded to unfamiliar situations, what they do when they don't know something. It means designing more of the work itself as a development environment- giving people experience that stretches them in the right direction, with the right support, rather than experience that confirms what they already know how to do. The 70-20-10 framework- which places most development in experience, some in relationships, and a smaller proportion in formal learning- has more relevance now than when it was first proposed, because formal skills training dates and on-the-job development of judgment and agility does not.

And it means being honest about what AI is for in the development context. Used as a thinking partner- something that challenges your reasoning, offers alternative framings, pushes back on your assumptions- it is one of the most powerful development tools available. Used as a replacement for thinking, it is the opposite. The difference is in the design, and in whether the organisation has been clear about what it is actually trying to build.

The organisations that thrive over the next decade will not be the ones with the most comprehensive skills framework. They will be the ones that understood, early enough, that the ability to learn is the skill- and built everything around that.

Alice Veitch is a Learning & Development Leader with a background in behavioural science. She has led learning strategy across EMEIA and is a respected disruptor in the field.