The 62% Premium
Companies are paying a premium to the workers AI makes more expert and erasing the on-ramps for the workers it makes more replaceable,
Welcome to Silicon Sands News, the go-to newsletter for investors, senior executives, and founders navigating the intersection of AI, deep tech, and innovation. Join ~35,000 industry leaders across all 50 U.S. states and 117 countries, including top VCs from Sequoia Capital, Andreessen Horowitz (a16z), Accel, NEA, Bessemer Venture Partners, Khosla Ventures, and Kleiner Perkins. Our readership also includes decision-makers from Apple, Amazon, NVIDIA, and OpenAI.
This week, we explore the labor-market split the displacement headlines keep missing: AI is making one track of workers more valuable and the other more replaceable, and the gap between them now decides which companies grow.
Let’s Dive Into It….
Key Takeaways
For VCs and LPs
Which track a company hires into is the new leading indicator: PwC’s analysis of more than a billion job ads finds roles “professionalized” by AI growing twice as fast, with 42% faster wage growth since 2021, than roles AI “democratizes.” A portfolio company stacking senior, judgment-heavy roles is buying durable value. One staffing up cheap, deskilled execution is renting a margin that competitors can rent too. The org chart is the tell, more than the model.
The headcount-growth story is real, and it is survivorship bias: PwC’s company data covers large firms that survived from 2018 to 2025, so the 52% headcount growth at the most AI-exposed firms versus 36% at the least AI-exposed measures winners consolidating share, not a clean one-to-one return on AI spend. Test whether the growth is due to causation or just survivorship, and ask what the failed cohort looked like.
The returns concentrate in a thin top slice: among the most AI-exposed companies, the top 20% posted 163% productivity growth, compared with 34% for the group. That is a superstar distribution, in which a thin top slice captures most of the gains. Price the category accordingly and back the firms with the structural traits of that top slice.
The broken entry pipeline is a liability you can price today: Stanford’s payroll analysis shows employment for 22-to-25-year-olds in the most AI-exposed jobs down 13% since late 2022, while their older colleagues held steady. A company that stops hiring juniors books a margin now and a senior-talent shortage in 2035. Diligence is how it builds judgment, the part that the cost story hides.
Workforce intelligence is the picks-and-shovels position: demand for agentic-AI skills jumped more than 280% in a single year, and running hybrid teams of people and agents needs tooling that maps skills and redesigns work. The category is early, buyers are forced, and spending is non-discretionary once a company commits to agents.
For Senior Executives
Sort every role by professionalized or democratized before you set the budget: PwC’s two-track lens tells you where to invest in expertise and retention and where to redesign work and open mobility instead. Professionalized roles demand new skills at twice the rate of democratized roles, so the two tracks require opposite talent plans. The framework is the plan.
The organization is the lever that moves AI impact: Microsoft’s analysis of trillions of work signals finds organizational factors such as culture, manager support, and talent practices drive about two-thirds of AI’s impact versus one-third for individual capability, and only 19% of AI users sit in the zone where capability and readiness reinforce each other. Plugging AI into today’s workflows buys efficiency at a cost multiple that rarely pencils; the gains come from redesigning the work and resetting incentives toward the org you want to become.
Do not cut past what the technology can actually do: BCG models 50% to 55% of US jobs being reshaped over the next two to three years; most of those roles survive in changed form. Companies that cut beyond AI’s real capability lose productivity and watch institutional knowledge walk out the door. Reshape roles deliberately and keep the people who hold the context.
Your entry pipeline is the quiet risk on the balance sheet: junior roles now carry senior expectations. PwC finds that traditionally senior skills account for 52% of new skill demand in the most AI-exposed entry-level jobs, compared with 7% in the least AI-exposed. Build apprenticeship, mentorship, and simulation now, or you starve your own management bench a decade out.
Pay the premium on purpose: the 62% wage premium for AI skills, reaching 118% in consumer markets, is a retention cost the budget has to carry. The scarce input is the worker who pairs AI fluency with judgment. Budget for that worker before the market reprices them.
For Founders
The professionalized track is a small-team weapon: AI amplifies a handful of experts, so a lean team can now carry work that used to require a department. Build the company that turns five strong operators into the output of fifty, and make that your pitch.
Make the pitch about new capacity: the firms winning with AI use it to enter new markets and expand, then hire senior talent to run the expansion. Position your product as a growth lever that a buyer can point at new revenue, because efficiency stories get value-engineered to zero.
Build for the broken rung: Brookings finds nearly 11 million workers without four-year degrees in “Gateway” jobs, with almost half the pathways to higher-wage work now exposed to AI. The eroding on-ramp is a market opening for tools that train judgment and scaffold junior workers into roles that used to take years to reach.
Sell workforce intelligence into a forced buyer: with agentic-AI skill demand up more than 280% in a year, enterprises have to map skills and manage teams of people and agents they were never built to run. The pain is acute, new, and budgeted.
Instrument the human-intensive skills: PwC finds the new tasks added to AI-exposed roles lean about 2.5 times more on capabilities like empathy, judgment, and creativity. Products that measure or develop those capabilities sit on the growth side of the split, where the wage premium and the hiring are.
Employment Dichotomy
A worker who can use AI well now earns 62% more than an otherwise identical worker who cannot use AI. A year ago, that gap was 57%. Two years ago, it was about 25%. PwC drew the number from more than a billion job ads across 27 countries, and it is the steepest premium any single skill set has ever commanded.
The same technology that is widening that gap is pulling the bottom rung off the ladder workers used to climb to reach it. In the occupations most exposed to AI, employment among 22- to 25-year-olds has fallen by 13% since late 2022, even as employment has held steady for their older colleagues doing the same work.
Both numbers come from the same force, and together they describe a labor market dividing into two tracks. One track pays more and grows faster while demanding sharper human judgment. The other gets easier to do and cheaper to staff. For investors, which track a company hires into is becoming a leading indicator of whether it compounds. For operators, talent strategy is now the load-bearing part of AI strategy.
For three years, the story was displacement. We waited for the layoffs that would mark the real arrival of the AI era. The 2026 data tells a stranger story. The companies extracting the most from AI are hiring faster, paying more, and rebuilding their org charts to capture what the technology opens up. The displacement is real, but it is hitting one specific place on the ladder, and the growth is real, but it is concentrating in one specific kind of company. Miss the split, and you misread both.
The two tracks, defined by expertise
PwC’s 2026 Global AI Jobs Barometer sorts the labor market by a single question: Does AI raise or lower the expertise a job requires?
When AI takes the routine work and leaves the hard parts to people, it professionalizes the role. PwC’s examples are radiologists and recruiters. An AI agent screens a thousand resumes, and the recruiter spends the day on the parts that need a human touch: reading the room, negotiating the package, and closing a contested candidate. The recruiter’s job survived. It moved up the skill ladder.
When AI takes the expert work and leaves the routine, it democratizes the role. The skill barrier drops, more people can do the job, and the wage pressure follows. By PwC’s count, 52% of jobs are being democratized, 22% are being professionalized, and 26% are at low exposure.
The two tracks pay and grow on different terms, and the gap is widening. Since 2018, job postings for professionalized roles have increased by 39%, compared with 17% for democratized roles. Salaries tell the same story: professionalized pay has risen 37% since 2021, compared with 26% for democratized, a 12-point gap PwC leads with. And the skill demands are pulling apart fastest of all, with professionalized roles asking for new capabilities at twice the rate.
That divergence is the whole thesis in one shape. AI works as a sorting mechanism, and the sort runs on expertise: it pushes some jobs up the skill curve and holds others down.
Why the AI leaders keep hiring
If automation simply replaced labor, the companies most exposed to AI would be shedding workers. Instead, they are adding them faster than their peers.
PwC finds that headcount at the most AI-exposed companies has grown 52% since 2018, compared with 36% at the least AI-exposed, and the gap has widened every year. Wages followed, up 24% at the exposed firms, compared with 17%. The companies leaning hardest into AI are hiring more people and paying them more.
The reason is in how they deploy it. The firms getting the most from AI use it to redesign their businesses, enter new markets, and treat cost-cutting as the smaller prize. Productivity gains turn into expansion, and expansion needs people, mostly senior ones, to run it.
The gains do not spread evenly. The most AI-exposed companies posted 34% productivity growth against a 2018 baseline, but the top 20% of that group hit 163%, roughly five times the average. A small set of superstar firms is capturing most of the value, and those same firms are the ones hiring fastest.
For an investor, that is the signal worth more than the category average. The returns from AI are concentrating, and the concentration shows up in who is growing headcount and who is not.
The survivorship asterisk
Before anyone celebrates the hiring numbers, read the methodology, because it changes what they mean.
PwC’s company analysis covers larger, formalized firms that survived from 2018 through 2025. It excludes startups, smaller businesses, and companies that failed or exited over those seven years. We are looking at a sample of winners.
That reframes the headcount story. These superstar firms did not necessarily grow because they adopted AI. They are structurally sound, well-capitalized organizations that were positioned to adopt AI effectively and grow their share while doing so. Their hiring reflects market consolidation as much as any one-to-one AI dividend. The causation runs at least partly in the opposite direction to how the headline reads.
It does not make the finding wrong. It makes it conditional. AI is a growth lever for companies already built to pull it. For everyone else, the same technology can still hollow out the middle. The honest version of the bullish case is narrow: among firms that survive and scale, AI exposure tracks with hiring. The firms that did not survive are not in the dataset, so there is no basis to argue otherwise.
The bottom rung is breaking
Cross the enterprise hiring data against the research on early careers, and a hard picture comes into focus. The jobs AI is creating are not landing evenly across the seniority ladder.
The technology is precisely good at the routine, execution-heavy tasks that used to be how junior people learned the business. So companies need fewer 22-year-olds. Stanford’s payroll analysis already shows it: a 13% relative decline in employment for early-career workers in the most AI-exposed occupations, concentrated where AI automates rather than augments. For young software developers, the drop runs steeper.
The damage spreads beyond the entry job itself. Brookings finds more than 15 million workers without four-year degrees in highly AI-exposed jobs, nearly 11 million of whom are in “Gateway” occupations that have historically enabled people to build skills and climb into higher-wage work. Almost half the pathways from those Gateway jobs to better ones are now exposed to AI. The on-ramps are being automated, not just the destinations.
This is the broken pipeline, and it is an existential problem with a long fuse. For decades, the path was clear: enter at the bottom, do the routine work, absorb the context, develop judgment, and rise. If companies stop hiring at the bottom because AI is cheaper, who will have the deep contextual knowledge to be the senior leaders of 2035?
The entry-level jobs that remain are changing, too. Because AI handles the routine, the expectations for juniors have jumped. PwC finds that traditionally senior skills now account for 52% of new skill demand in the most AI-exposed entry-level roles, compared with 7% in the least AI-exposed. Entry roles that absorb those senior skills are growing 35%, while the rest decline. We are asking 23-year-olds to supervise and correct AI output, which takes expert judgment, before they have done the underlying work long enough to have any.
BCG puts a number on the scale of the reshaping underneath all of this: 50% to 55% of US jobs will be reshaped over the next two to three years, with a large share of entry-level tasks absorbed by machines. Most workers keep their jobs and face radically different expectations for how they work. The lower rungs of the ladder are being sanded off.
What educators and new graduates can do
The diagnosis is grim, but the gap is closeable, and it falls to two groups to close it: the institutions training young workers and the young workers themselves.
Start with educators, because the curriculum is aimed at the wrong target. For decades, schools taught students to produce the work, and the first job taught them to judge it. AI inverted that order. The routine production is the part machines now do, and the judgment, the part that used to take years on the job, is the part employers want on day one. PwC says it plainly: companies and educators have to rebuild the way they mentor and scaffold early careers so juniors can demonstrate senior-level skills far sooner.
In practice, that means putting students in the supervisor’s seat from the start. Hand them AI that produces the first draft, the first analysis, the first model, and grade them on whether they can catch what it got wrong, push it where it fell short, and stand behind the result. Teach AI fluency as a baseline, then spend the real hours on what AI cannot do: framing a problem, reading a stakeholder, exercising judgment under ambiguity, and leading a room. Wrap it in project work that looks like the job, with AI in the loop, a real deliverable, and a real critique, so a graduate arrives having already run the loop that used to take three years to learn. The schools that wire employers into that loop, rebuilding the apprenticeship that automated entry work no longer provides, will place graduates that the others cannot.
Now the 22-year-old, who has the harder task and less time. The first move is to become the person who commands AI. The premium the market is paying, that 62% number, rewards the person who pairs AI fluency with judgment, who can take a problem, point a system at it, and stand behind the answer. Build that combination on purpose.
Aim for the professionalized track, the work where AI raises the value of your judgment: the decisions, the persuasion, the responsibility a machine cannot carry. Learn to deploy agents, the systems that take action, because that is where demand is moving and where the next premium will sit. Get close to work that compounds context, the kind that builds the domain judgment a senior role runs on, even while AI handles the routine around you. The skill that pays now is directing AI and owning the outcome, because your competition for the job is whoever does that better than you.
The real bottleneck is the org
The technology is ready. The senior workers are increasingly ready. The thing that is not ready is the organization.
Microsoft analyzed trillions of work signals and surveyed 20,000 AI users, and the headline is blunt: people are further along than the companies they work for. Organizational factors like culture, manager support, and talent practices account for about two-thirds of AI’s measured impact, against one-third for individual capability. What a company builds around its people matters more than twice as much as the people.
The distribution is stark. Only 19% of AI users sit in the “Frontier” zone, where individual capability and organizational readiness reinforce each other. About half are still in the messy middle. And 10% are blocked: skilled workers stuck in companies that have not caught up to the capabilities their own people already have. Treat the soft survey numbers as direction rather than precision, but the direction is consistent with three years of independent research. The constraint moved from the model to the management.
Two moves separate the companies that capture value from those that buy licenses and wait. The first is incentives. Most organizations reward the behavior they already have, then act surprised when AI changes nothing. The companies pulling ahead reset incentives toward the org they want to become, so the people closest to the work are rewarded for redesigning it.
The second is the workflow itself. Plugging AI into an existing process is the common move and the expensive one. It buys 5x the efficiency on a task but costs 10x to run it, and at scale, that math does not pencil. The value shows up when you rebuild the workflow around what AI can actually do. The bolt-on version is the one that disappoints.
I have watched this from the inside, so weigh it as the view of an interested party. We started the way most teams do, with off-the-shelf coding tools like Cursor and Claude Code, and got the efficiency bump everyone gets. Then we built our own agentic workflow from the work outward. It now does more than anything we could buy, and it is deterministic where determinism matters. We don’t focus on agents as employees; we focus on agents as work that needs to get done. The advantage came from rebuilding the work, and that is the lesson the enterprise data keeps teaching at scale.
This is why demand is shifting from chatbots to agents. Skills tied to agentic AI grew more than 280% in a single year, from a rounding error to roughly 90,000 US job postings. The work is moving from talking to AI to deploying systems that take action within real workflows, and managing a hybrid team of people and agents is a job most org charts were never built to handle. Talent-intelligence platforms are racing to build tooling because the companies getting real value from AI are far more likely to deploy agentic systems.
How to read the split
The funding tells you where the conviction is. The split tells you where the returns will land.
The bullish version of the AI-jobs story is true but conditional. Among companies built to scale, AI exposure tracks with hiring, higher wages, and a superstar tier capturing most of the gains. The bearish version is also true, in a different place: the entry pipeline is breaking, and the workers who used to climb through Gateway jobs are watching the rungs disappear. Both are happening at once, and a portfolio or an org that sees only one is mispriced.
For capital, the discipline is to underwrite the cause rather than the correlation. Headcount growth at an AI-exposed company can mean it is a structurally sound winner consolidating share, or it can be survivorship noise. The company worth backing is the one building toward the professionalized track in a domain where judgment compounds, paying the premium for the workers who pair AI fluency with that judgment, and rebuilding the apprenticeship that its own future depends on. The fragile bet is the company selling efficiency, cutting its juniors, and booking a margin today against a senior-talent shortage it will feel in a decade.
Most of the market is not yet reading the data this way. That gap is the edge. The split is in the open, measured across a billion job ads, and the diligence that uses it is still rare.
Let’s Wrap This Up
The displacement story was never quite right, and the replacement story taking its place is more dangerous because it is half-true. At the top, superstar companies are paying a 62% premium and hiring senior talent faster than anyone. At the bottom, the ladder that fed them is losing its rungs.
Both are true at once, and the companies that win the next cycle will be the ones that hold them together: paying for judgment where AI makes it scarce, and rebuilding the on-ramp where AI is automating it away. The technology is ready, and the senior workers are ready. The only open question is whether your organization is built to train the leaders it will need in 2035, starting with the juniors you are deciding whether to hire today.
Disclosure and disclaimer. The author, Dr. Seth Dobrin, is the founder of ARYA Labs, referenced once above with disclosure. The views expressed are his own. Silicon Sands News is an independent publication. Nothing here is investment, legal, or financial advice, and it should not be relied on as a basis for any investment decision. Companies and research referenced are cited to primary sources for the reader’s own evaluation. Each chart cites its sources in its footer, and the underlying data for every chart is available from the chart itself.