PwC's 2026 Barometer named the two-track labor market: AI lifts expertise and pay in some jobs and commoditises others. See which track your role is on.
TL;DR: PwC's 2026 Barometer split AI-affected work into two tracks: professionalised roles where AI raises your expertise and pay, and democratised roles where AI lowers the skill bar and flattens wages. Aim to be the human who directs AI, not the one competing with it.
The two-track labor market is the split PwC named in its 2026 Global AI Jobs Barometer between two kinds of AI-affected work. On the winning track sit "professionalised" jobs, where AI handles the basic tasks and humans keep the expert judgment, so demand and pay climb. On the losing track sit "democratised" jobs, where AI does the expert work and the person is left with the routine parts, so more people can compete for the role and wages stall.
About a quarter of advertised jobs globally are professionalised and about half are democratised, based on PwC's analysis of more than a billion job ads. The rest have low exposure to AI. This piece defines both tracks plainly, helps you work out which one your own role is on, and covers the evidence-based ways to move toward the winning side.
PwC's June 2026 report puts a name to something a lot of workers already sensed: AI is not simply adding or destroying jobs, it is changing what expertise each job requires. PwC calls the outcome a "two track labour market," using British spelling, so you will see the term written both ways, two-track labor market and two-track labour market. The distinction it draws is about direction: does AI push the expertise bar of your job up or down?
The idea of a split labor market is older than AI. Economists Peter Doeringer and Michael Piore described a dual labour market of primary and secondary segments back in the early 1970s. PwC's fresh contribution is the AI-specific axis of professionalised versus democratised, not the general notion of a divide.
The headline finding is blunt. PwC reports that the 22% of jobs being professionalised are growing twice as fast as the 52% that are democratised, with 42% higher wage growth since 2021. That wage figure is a relative gap, not 42 extra percentage points: professionalised salaries grew 37% since 2021 while democratised salaries grew 26%, and 37 is roughly 42% more than 26.
The barometer is built on more than one billion online job ads across six continents. The press release, dated 15 June 2026, pins the scope at 27 countries and territories. That scale is why the report gets cited as a bellwether rather than a single-country snapshot.
Joe Atkinson, PwC's Global Chief AI Officer, framed the shift as structural. "Across the global economy, we're beginning to see a new divide emerge between different models for talent and value creation," he told IT Pro. The report backs this with two supporting signals worth remembering.
First, skills are churning fast. PwC found that the skills required for the most AI-exposed jobs are changing twice as fast as in the least exposed roles, while new tasks that lean on empathy, judgment and creativity are added 2.5 times faster. Second, AI skills now pay: workers with them command a 62% wage premium, up from 57% the year before.
PwC defines the two tracks by how AI changes the expertise a role demands. Professionalised jobs are "reshaped by AI to demand more expertise," with examples including radiologists, employment recruiters and air traffic controllers. Democratised jobs are "reshaped by AI to demand less expertise," and PwC's own examples here are software developers, loan officers and finance managers.
The report offers a clean analogy. Think of a lawyer whose AI summarises documents while the human argues the case: that is professionalisation, because the expert layer stays with the person. Now think of an inventory clerk whose AI manages the inventory while the human moves stock: that is democratisation, because AI took the expert part and left the routine.
Here is the twist that catches people out. "Democratised" sounds like good news, but for the individual worker it is usually the losing track. When AI lowers the skill barrier, more people can do the job, so wages stagnate even when openings rise.
Notice that PwC lists software developers as democratised. That directly contradicts the pile of "AI-proof jobs" listicles that still tell people to learn to code because programming is safe. Under this data coding is a role where AI does more of the expert work, so the skill that stays valuable is closer to AI fluency than to any one programming language.
The clearest winners-and-losers signal in the report is what PwC calls a superstar effect. The most AI-exposed firms as a whole grew productivity 33.5% since 2018, but the top fifth of them, the superstars, grew 163%. Productivity here means revenue per employee, so the gains concentrate at the frontier rather than spreading evenly.
For workers, the trap is mistaking more openings for a better deal. PwC's example is child care services managers, a democratised role: job listings have more than doubled since 2019, up 111%, while wages have grown only 8%. More postings and flat pay is what the losing track looks like in practice, which is why "there are plenty of jobs in my field" is not proof you are winning.
Independent coverage lines up with the pattern. Euronews reported the same roughly double job growth and 42% faster salary growth for the professionalised track. The winning track is not a specific job title, it is a position: being the judgment and direction layer that sits on top of AI.
You do not need PwC's dataset to place your own role. Ask four plain questions and answer them honestly.
If AI is taking your grunt work and you are increasingly paid for judgment, taste and direction, you are trending professionalised. If AI is doing the skilled part and you are left assembling or supervising its output, you are trending democratised, even if postings are plentiful. Sitting with that answer can be uncomfortable, and if it is stirring real dread, our guide to AI job anxiety is a calmer place to start than doomscrolling.
The honest answer is that no job is truly AI-proof, and "AI-proof" is the wrong target. PwC's low-exposure category, which includes chefs, construction workers and mechanics, is the most insulated from AI today. But low exposure also tends to come with a smaller AI wage premium, so safety and upside pull in different directions.
The more durable answer is not a job title, it is a stance. The roles pulling ahead are the ones where a human directs AI and supplies the judgment AI cannot, which is why PwC's own advice centers on becoming that person rather than hiding from the technology. Remember the developer point: PwC classifies software development as democratised, so the popular "learn to code because it is future-proof" advice runs backwards against this specific data.
It is worth staying honest here, because the two-track story can read as more dramatic than the current data supports. The Yale Budget Lab found no clear economy-wide disruption as of early 2026. Its director, Martha Gimbel, put it plainly: "it just doesn't seem like there's major macroeconomic effects here."
Where the signal does show up is at the bottom of the ladder. Stanford's Canaries in the Coal Mine study found a 16% relative decline in employment for 22 to 25 year-olds in the most AI-exposed occupations, while older and less-exposed workers held steady or grew. PwC's own CEO survey adds that 49% of chief executives expect AI to cut junior hiring over three years, versus 12% for senior hiring.
PwC calls this "seniorisation": traditionally senior capabilities now make up 52% of the new skills required for AI-exposed entry-level jobs, against just 7% for the least-exposed entry roles. Entry roles that have been seniorised this way grew 35%, while globally the number of entry-level jobs in highly AI-exposed roles has flatlined. This is the mechanism behind the broken skills ladder: the bottom rung that used to train people is being automated away.
For the wider backdrop, the World Economic Forum projects 170 million new roles and 92 million displaced by 2030, a net gain of 78 million jobs. Churn is real, but net collapse is not the forecast.
PwC ends with five steps for workers, and they double as a practical plan. Move toward roles that AI makes more expert rather than less, and seek out pioneering, AI-first teams where directing the technology is the job. Above all, learn to command AI as a tool and partner, because in PwC's framing you will not lose your job to AI but to someone who knows how to use AI.
The last two steps are about staying human on purpose. Build the human-intensive skills the report keeps flagging: creativity, people skills, leadership, judgment and the ability to navigate ambiguity. And if you are early in your career, develop these soft skills AI cannot replace fast, since the entry rung is exactly where AI is biting hardest.
This is where deliberate practice beats reading another trend piece. GPTnius pairs you with an AI mentor built to grow exactly the two capabilities that define the professionalised track: sharper judgment and the skill of directing AI well. If you want a structured start rather than a vague resolution, try your AI mentor free and calibrate it to your role and your track.
It is PwC's 2026 framing for how AI is splitting work into two paths. Professionalised jobs, about 22% of roles, use AI to handle basic tasks while humans supply expert judgment, so pay and demand rise. Democratised jobs, about 52%, use AI to do the expert work and lower the skill bar, so wages stagnate even as openings grow.
Professionalised jobs are reshaped by AI to demand more expertise, with examples like radiologists, recruiters and air traffic controllers. Democratised jobs are reshaped to demand less expertise, and PwC lists software developers, loan officers and finance managers here. The catch: democratised sounds positive but is usually the losing track, because a lower skill bar means more competition and flatter wages.
No job is truly AI-proof. PwC's low-exposure roles, such as chefs, construction workers and mechanics, are the most insulated, but low exposure also tends to mean a smaller AI wage premium. The more durable position is not a title but a stance: being the human who directs AI and supplies judgment. Note that PwC classifies software developers as democratised, so 'learn to code because it is safe' runs backwards here.
Not quite. The dual labour market idea of primary and secondary segments was developed by economists Peter Doeringer and Michael Piore in the early 1970s. PwC's contribution is the AI-specific axis: whether AI raises expertise (professionalised) or lowers it (democratised). PwC uses British spelling, so you will see both two-track labor market and two-track labour market in coverage.
At the economy-wide level, not much yet. The Yale Budget Lab found no major macroeconomic effects in early 2026. The clearest signal is at entry level: Stanford found a 16% relative employment decline for 22 to 25 year-olds in the most AI-exposed jobs, and 49% of CEOs told PwC they expect AI to cut junior hiring. The split is real but concentrated, not universal.