Cognitive
How you break a problem apart, sequence work, put an idea into someone else’s head, and change your mind. The domain most exposed to generative AI — and, paradoxically, the one where the human premium is moving up the stack, from producing analysis to judging it.
Critical thinking 4
Planning & ways of working 3
Communication 4
Mental flexibility 5
Taking a vague, data-poor question and turning it into a set of components that don’t overlap and don’t leave gaps; forming a falsifiable hypothesis; then sequencing the analysis by value of information rather than by what’s easy to pull.
They draw the tree before they open the spreadsheet. They can tell you, unprompted, what evidence would prove them wrong. When you hand them a contradicting fact at minute 30, they redraw a branch instead of defending the answer.
It is not intelligence, speed, or being well-read — plenty of very clever people solve unstructured and get lucky. It is not slide craft. It is not “having the answer”: someone who arrives with the conclusion and reverse-engineers support is demonstrating the opposite skill.
A 45–60 minute live case on a problem they’ve never seen, with deliberately missing data. Score the process on five anchored dimensions: clarified the question / decomposition is non-overlapping / prioritised by impact × effort / stated a falsifiable hypothesis / updated on new evidence. Two independent raters, then check they agree. McKinsey’s own Solve assessment does a gamified version of exactly this.
- Change the artefact, not the attitude. For twelve weeks, no analysis begins without a one-page issue tree signed off by a peer. Skill follows the mandatory artefact.
- Force the falsifier. Every recommendation carries a sentence beginning “This is wrong if…”. It is the cheapest debiasing intervention that exists.
- Review the tree, not the answer. If managers only ever critique conclusions, people optimise for confident conclusions.
- Re-test cold at week 0 and week 12 on a fresh case with different raters. If the rater score hasn’t moved, the problem was never training — it was that nobody has time to think.