Is AI pulling every job toward one shared skill set, or splitting work into a tight tech core and everything else?

This explores whether jobs heavily exposed to AI ask for a different set of skills than other jobs, and whether AI is pushing every occupation toward one common toolkit. The clearest evidence in the collection says no to the common toolkit. Job-ad data from ten countries shows AI skill demand clustering inside STEM occupations, around Python, SQL, machine learning and data analysis. Non-technical occupations don't move toward that core. They drift further from it Is AI creating common skills across jobs or deepening divisions?. So the gap isn't simply that AI-exposed jobs need more AI skills. It looks like a split, with a tight technical core on one side and a widening, more varied set of jobs on the other.

Where AI is actually used is a different question from where it could be used. When you track tasks workers have handed off to AI as part of their regular workflow, rather than casual chatbot use, the uptake concentrates in information-heavy work and follows what the technology can actually do. It doesn't match older predictions that 'routine' jobs would be automated first. Wage patterns even reverse among workers with advanced degrees Where have workers actually delegated tasks to AI?. Exposure also lands differently depending on who holds the job. In male-dominated fields it mostly reaches high-skilled, high-paid workers. In female-dominated fields it spreads evenly across skill levels, so lower-paid women are exposed with fewer resources to adapt Does AI exposure hit low-wage workers harder in some fields?.

Outside the technical core, the skill AI-exposed employers reward looks less like engineering and more like being able to say you work with AI. In a hiring experiment with 1,725 recruiters, listing AI skills raised interview invitations by 8 to 15 percentage points across occupations. A certificate added only a little over simply claiming the skill, which suggests recruiters reward the label without checking competence Do AI skills help candidates get more job interviews?. The same signal partly offset penalties for being older or not having a bachelor's degree. The effect was strongest for office assistants and weaker for graphic designers Can AI skills help older or less-educated job candidates?. In other words, how much an AI skill is worth depends on the occupation.

The less obvious finding is about which skills last. Being exposed doesn't always mean being displaced. When AI touches only a few of a job's tasks, workers can shift their time to the remaining tasks, which softens the effect on employment Does concentrated AI exposure enable workers to adapt and reallocate?. That makes the skills AI doesn't cover the valuable part of an exposed job. There is a catch, though. Abilities boosted by AI tend to work like an exoskeleton: they disappear when the tool is taken away Does AI assistance build lasting skills or temporary abilities?. Anthropic's own engineers report big productivity gains, yet they can fully delegate only a small share of their work. They also worry that relying on the AI erodes the hands-on practice they need to catch its mistakes Does AI assistance erode the skills needed to oversee it?. So the hardest skill to see, and maybe the most important one in exposed jobs, is being able to judge AI output without the AI.

A note on what's missing: apart from the ten-country vacancy study, the collection doesn't directly compare skill requirements in exposed and unexposed sectors. Most of the material covers exposure, hiring signals or productivity. The picture above is pieced together from those adjacent findings, not taken from one head-to-head comparison.