PwC's 2026 Global AI Jobs Barometer, released June 15, puts a 62 percent average wage premium on roles requiring AI skills — up from 57 percent a year earlier and 25 percent the year before that. But that headline number isn't the finding that should change how you hire. Underneath it, built on more than a billion job listings across 27 countries, PwC split the labor market into two tracks. Which track your open roles fall into is a better predictor of your hiring outcomes this year than your budget is.
The report calls the two tracks "professionalised" and "democratised." Professionalised roles, like radiologists and recruiters in PwC's examples, are ones where AI automates the routine parts of the job and pushes the remaining work toward judgment, expertise, and things a model still can't decide on its own. Democratised roles, like IT service managers and medical secretaries in PwC's framing, are ones where AI just makes the job easier for someone with less training to do. Same technology, opposite effect on the person doing the work: one track gets more valuable, the other gets more replaceable.
The gap between them isn't subtle. Professionalised roles are seeing twice the job growth of democratised ones, and 42 percent faster salary growth. Jobs requiring AI-specific skills grew 69 percent year over year against 9 percent for the overall market, roughly eight times faster and nearly double the 2024 count.
PwC reports a second divide too — not between roles, but between companies. My read is that the two are related. Joe Atkinson, PwC's global chief AI officer, framed it this way: "Across the global economy, we're beginning to see a new divide emerge between different models for talent and value creation. The companies seeing the greatest returns on AI are using it to amplify human expertise, accelerate innovation and create entirely new sources of value. As a result, they are pulling further ahead on productivity and growth than companies that focus primarily on automation."
Pete Brown, PwC's global workforce leader, added the harder part for anyone building a team: "The traditional relationship between experience and expertise is changing. AI is removing some of the routine work that once acted as an apprenticeship, while increasing demand for judgement, leadership and adaptability much earlier in careers."
Sit with that second quote, because it's the one hiring managers keep skipping past. The apprenticeship model most engineering orgs run on, where a junior does the routine work, absorbs judgment over a few years, and becomes the senior nobody has to double-check, depends on the routine work existing long enough to teach something. PwC's data says that window is closing faster than most org charts have adjusted for. Junior hires now spend less time on rote implementation and more time expected to exercise judgment that used to take three years to build. Most companies didn't build a faster path to get them there; they just deleted the first three years and kept year four's expectations.
The premium data shows who's already figured this out. Consumer markets carry a 118 percent AI-skills wage premium — nearly double the 62 percent average — while government and public sector sits at 16 percent. That's not an industry quirk; it's the same divide showing up in what employers will pay. Sectors moving fastest toward professionalised roles are bidding aggressively for the judgment layer. Sectors still hiring for democratised roles aren't, because they don't need to yet. They will.
Here's where this stops being a labor-economics story and starts being a job-description problem. Most companies are still writing reqs for the democratised version of a role, with its years of experience, tool proficiency, and checklist of routine tasks, when the job they actually need filled has quietly become the professionalised version: less about executing the workflow, more about knowing when the AI's output is wrong and what to do about it. You can't source for that with the old req. A candidate who's spent two years being the person who checks the model's work, argues with it, and catches what it misses is a different hire than someone who's spent two years doing the task the model now does faster. Same job title on LinkedIn. Not the same candidate.
If you're budgeting headcount for next year, the question isn't "can AI do this role's tasks" — PwC's data says the answer for most technical roles is already partially yes. The question is whether the role you're hiring for sits on the professionalised or democratised side of that split. That 62 percent premium isn't paid out evenly across the market; it's concentrated on one side of a line most job descriptions haven't noticed yet. Write the req for the track you need. Not the one you used last time.
VC5 Consulting finds the engineers whose value went up when AI arrived — the ones doing the judgment work PwC's data says is getting professionalised, not automated. If your reqs still describe the job AI already ate, let's talk.