Job postings suggest AI skills are splitting the workforce in two: a technical core, and everyone else moving away from it.

This explores how the demand for AI skills, and AI's reach into work more broadly, differs from one occupation to another. The short version from the corpus is that AI demand doesn't spread evenly across jobs. Vacancy data from ten countries show employers' requests for AI skills clustering in STEM roles around a shared toolkit of Python, SQL, machine learning and data analysis. Non-technical occupations aren't gradually adopting that toolkit. They are moving away from it, so the labor market splits into a technical core and everyone else rather than converging on one universal set of 'AI skills' Is AI creating common skills across jobs or deepening divisions?.

What employers ask for and where AI actually gets used are two different maps. When researchers track where workers have built AI into structured workflows, rather than just chatting with a model, it concentrates in information-heavy jobs and follows what the technology can actually do. That pattern doesn't match older predictions that automation would hit 'routine' jobs first. The wage pattern also flips at the advanced-degree level Where have workers actually delegated tasks to AI?. Gender adds a less obvious layer. In male-dominated occupations, AI exposure falls mostly on high-skilled, well-paid workers. In female-dominated occupations it spreads evenly across all skill levels, so lower-paid women face substantial exposure with fewer resources to adapt Does AI exposure hit low-wage workers harder in some fields?.

The finding you might not expect is that the amount of exposure isn't the whole story. How that exposure is spread across a job's tasks matters too. Firm-level data from 2010 to 2023 show that higher average exposure does reduce demand for labor. But when exposure is concentrated in a few tasks, workers can shift toward the tasks AI doesn't touch, and the net effect on employment ends up modest Does concentrated AI exposure enable workers to adapt and reallocate?. So a job heavily affected in one corner may be safer than a job lightly affected everywhere. Employers vary as well: firms with more AI exposure replace online freelance workers with AI tools faster and more cheaply than other firms. That suggests firms build in-house AI capability that pays off at scale, rather than the technology spreading at the same pace everywhere Do firms substitute labor for AI at different rates?.

On the hiring side, AI skills pay off broadly even though actual demand is narrow. In an 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, so recruiters seem to reward the signal without checking real competence Do AI skills help candidates get more job interviews?. That fits signs of an escalating loop in hiring, where applicants use AI to beat filters and employers add more filtering in response Are job applicants and employers locked in an escalating AI arms race?.

One caution on any 'which jobs will AI affect' map: much of it rests on capability measures that may not reflect real work. An analysis of 960 real occupational workflows found that AI agents excel at contest-style benchmarks but struggle with long, multi-step professional tasks Why do agent benchmarks not predict real economic value?. The corpus is strong on these structural patterns but doesn't offer a detailed occupation-by-occupation breakdown. If that's what you're after, the vacancy and delegation studies above are the best places to start.