Is AI Replacing
Jobs or Creating
Them?
We spent two weeks in the research. The answer isn't reassuring — but it isn't a death sentence either. Here's what the data actually says.
92 Million Displaced. 170 Million Created. That's the Whole Story.
The question "Is AI replacing jobs?" is simultaneously the most asked and the most poorly answered question in technology right now. Apocalyptic headlines and dismissive optimism exist in equal measure — both distorting a picture that is genuinely complex, genuinely consequential, and genuinely different depending on who you are and what you do.
We went to the primary sources. The World Economic Forum's Future of Jobs Report. Dallas Federal Reserve wage research. Harvard Business School's labour market study. McKinsey Global Institute projections. BLS employment data. Here is what they collectively say: AI is simultaneously the largest job-displacement force and the largest job-creation engine in modern economic history — and both are happening at the same time, to different people, in different sectors.
The critical nuance that every headline misses: AI is replacing tasks, not jobs. Entire positions rarely vanish overnight. What vanishes first are the repetitive cognitive tasks inside existing jobs — and the roles that consist almost entirely of those tasks are the ones that disappear entirely.
"AI doesn't eliminate work — it eliminates tolerance for average performance. The workers most at risk aren't those whose tasks overlap with AI. They're those who refuse to use AI to do those tasks faster." — Dallas Federal Reserve, 2026
The Automation Risk Map — Where the Pressure Is Heaviest
Not every job faces the same pressure. The research is consistent: roles built primarily on codified, repetitive, rule-based cognitive tasks face the highest displacement risk. Roles that demand tacit knowledge — the kind you only gain through years of human experience, relationship-building, and contextual judgment — are far more resilient.
Based on Goldman Sachs research, McKinsey projections, and 2026 employment data, here is a data-backed risk ranking across key professions. The percentage reflects estimated automation exposure — not guaranteed job loss, but the proportion of role tasks that AI can already perform.
In March 2026 alone, 45,000 tech workers were laid off — with over 9,200 directly attributed to AI and automation. Amazon accounted for roughly 30,000 of those, flattening management layers and redirecting resources toward AI infrastructure.
A World Economic Forum survey found that 37% of companies expect to have replaced jobs with AI by end of 2026. But an important caveat from Harvard Business Review: many companies are laying off workers based on AI's potential, not its actual performance. They're betting that AI will handle the work — and in some cases, that bet isn't paying off yet.
The Other Side of the Ledger Nobody Talks About
In 2024, approximately 119,900 AI-related roles were added globally — far exceeding the confirmed AI-driven losses of the same year. Annual AI-related job creation is projected at 6 million for 2026 alone. The net impact trends positive, but the gains and losses hit different people in different sectors.
Professionals with specialised AI skills now command salaries up to 56% higher than peers in identical roles without those skills. The labour market isn't contracting — it's bifurcating. The question isn't whether there will be jobs. It's whether you'll be positioned on the right side of the divide.
LinkedIn data shows AI has already added 1.3 million new roles globally since 2023. Nurse practitioners are projected to grow 52% by 2033. Cybersecurity analysts face 32% growth through 2032. The jobs being created are better paid, more cognitively demanding — and unavailable to those who haven't upskilled.
Why History Says This Ends Better Than You Think
The most cited — and most important — historical analogy in any serious discussion of AI and employment is the ATM. When ATMs were introduced in the 1970s, economists predicted the extinction of the bank teller. The logic was airtight: machines could dispense cash faster, cheaper, and without sick days.
What actually happened is one of the most instructive case studies in economic history. The number of tellers per branch dropped significantly — but ATMs made it dramatically cheaper to open new branches. More branches meant more customer interactions. More interactions meant more tellers were needed — not fewer. The total number of bank tellers in the US roughly doubled over the following three decades.
The researchers at the Dallas Federal Reserve who documented this pattern in 2026 are seeing the identical dynamic emerging with AI: total employment is up 2.5% since ChatGPT's release. Wages in AI-exposed sectors are rising 16.7% — versus 7.5% nationally. The pattern is intact. The transition, however, is painful for those caught mid-career without the skills to adapt.
The ATM lesson isn't that automation is harmless. It's that automation reshapes work rather than eliminating it — and the workers who fare worst are those who wait for the reshaping to happen to them, rather than moving toward the new shape proactively.
Winners, Losers, and Everyone In Between
The impact of AI is not uniform across industries. Technology, healthcare, and renewable energy are net job creators in the AI era. Manufacturing and retail face the steepest structural headwinds. Finance and legal sit at an uncomfortable inflection point — bleeding entry-level roles while growing high-value specialisations.
This table represents cross-validated data from the BLS, McKinsey, WEF Future of Jobs 2025, and 2026 hiring trend analysis from LinkedIn Economic Graph.
20% of organisations will use AI to flatten their hierarchy by end of 2026 — eliminating over 50% of current middle management positions. The most dangerous position in any company right now is middle manager whose primary function is information relay. AI does that job better, faster, and for free.
Why Young Workers Are Feeling It Hardest — and What to Do
The data reveals a troubling generational asymmetry. Stanford University researchers Erik Brynjolfsson, Bharat Chandar, and Ruya Chen documented in early 2026 that the decline in employment in AI-exposed sectors is particularly pronounced for workers under 25. Employment for older, experienced workers has not declined.
The reason is structural. AI can replicate codified, textbook knowledge — the kind that entry-level employees rely on to demonstrate their value in the first years of their career. It cannot yet replicate tacit knowledge: the intuition, judgment, and contextual sensitivity that experienced workers develop over time. The result is that AI substitutes for entry-level workers while augmenting experienced ones.
The Dallas Federal Reserve confirmed that employment among workers aged 22–25 in AI-exposed roles has declined 13%. This is not happening through layoffs — it's happening through a collapse in the job-finding rate. Companies are hiring fewer entry-level staff because AI is absorbing the tasks those new hires would have performed.
66% of enterprises are reducing entry-level hiring due to AI — a pipeline shock that is hitting new graduates hardest. The traditional career ladder — enter at the bottom, learn the fundamentals, work your way up — is being removed from the bottom. The implication for how we educate, train, and onboard talent is profound and unsolved.
The counterpoint — and it matters — is that wages in AI-exposed occupations are rising faster than the national average. The jobs that remain in these sectors are better compensated precisely because they require the tacit, experiential knowledge that AI cannot replicate. The ladder hasn't disappeared. It's just missing its bottom rungs.
The Four Moves That Separate Those Who Thrive From Those Who Don't
The research consensus is clear: the workers most at risk are not those whose tasks overlap with AI. They're those who refuse to use AI to do those tasks faster and better. Adaptation isn't optional — but the path forward is more accessible than most people realise.
These four strategic moves are drawn from the intersection of labour economics research, hiring trend data, and real outcomes among workers who've navigated AI disruption successfully across sectors.
Start with one AI tool in your existing workflow. Ship something with it this week.
In every task ask: am I the one who defines the goal, or only the one who carries it out?
Identify one deeply human skill in your field. Invest in developing it deliberately.
Measure your output this month. Use AI tools to double it next month.
A content writer who uses AI to draft and edit produces 3x more than one who doesn't. The company doesn't fire the AI-assisted writer — they fire the one who takes three days to write what AI-augmented colleagues produce in one. The threat is not AI. The threat is refusing to use it.
What the Next Four Years Actually Look Like
The period from 2025 to 2030 is what labour economists are calling the critical transition window. The displacement and creation happening simultaneously today will resolve into a new equilibrium — but the path through is uneven, and the timing matters enormously for individual careers and companies alike.
The projections below synthesise WEF Future of Jobs 2025, McKinsey Global Institute, Oxford Economics, and Goldman Sachs research — calibrated against 2026 employment data that is already partially validating these forecasts.
By 2030, up to 30% of current US jobs could be automated. But 60% will have tasks significantly modified — not eliminated — by AI. And 59% of the global workforce will require upskilling or reskilling. The workers who start that journey now have a four-year head start on those who wait for the pressure to become undeniable.
The arithmetic ultimately trends positive. AI and automation could displace approximately 85–92 million jobs globally by 2030 — while creating 97–170 million new roles over the same period. The net is between 5 and 78 million new jobs depending on the pace of adoption. The question is not whether work survives. The question is whether you are positioned to access the work that does.
The Final Verdict
"AI is not replacing workers. AI is replacing workers who don't use AI. The distinction will define careers for the next decade."
Replacing: yes — 92 million roles globally by 2030.
Creating: yes — 170 million new roles over the same period.
The answer to "is AI replacing jobs or creating them?" is both, simultaneously, right now.
Which side of that equation you land on is still, largely, your choice.
