AI and the Future of Work Hits Young Workers First, and It Is Fixable

Last updated on October 22nd, 2026 at 12:53 pm

AI and the future of work is no longer a forecast. It is a payroll story, and the newest numbers say the damage so far is narrow while the promise is wide. Generative AI has not caused mass unemployment, but it is already thinning out entry-level jobs in the most exposed occupations, and the gains are flowing to workers and companies that learn to work with it.

The World Economic Forum still expects a net gain of 78 million jobs worldwide by 2030, yet in the United States employers have pinned 116,175 announced job cuts on AI in the first eight months of 2026. Both things are true at once, and this guide explains why.

Why I rewrote this article in 2026

I first published this essay in March 2025. That version ran to roughly 9,400 words and leaned on more than 30 books, from Klaus Schwab’s The Fourth Industrial Revolution to Nick Bostrom’s Superintelligence and Stuart Russell’s Human Compatible. The books gave me the big ideas. The trouble was the numbers. Most of them came from studies written before ChatGPT existed, such as the 2013 Oxford paper that estimated about 47% of US employment was at risk of computerisation.

Eighteen months later, we no longer have to guess. Payroll data, job postings, layoff announcements and company surveys now show what AI is actually doing to work. So I kept the questions the books taught me to ask and replaced the guesses with evidence. Where the evidence is thin or contested, I say so.

Will AI take my job?

Probably not outright, but it will very likely change what your job involves, and the risk is highest if your work is routine, text-based and done early in a career. The headline exposure numbers sound frightening until you read what “exposed” means. The IMF estimates that about 60% of jobs in advanced economies may be affected, and it expects roughly half of those to gain productivity from AI while the other half could see lower demand, lower wages or slower hiring.

The ILO’s 2025 update reaches a calmer conclusion. It finds that most jobs will be transformed rather than made redundant, because people are still needed to direct, check and finish the work.

Goldman Sachs puts a number on the downside. In its August 2025 analysis, it estimated that 6% to 7% of the US workforce could be displaced if AI is widely adopted, with a range of 3% to 14% depending on assumptions. It also expects unemployment to rise only about half a percentage point during the transition, as displaced workers move into other jobs.

Here is how I read those three sources together. A small share of jobs will disappear. A large share will be rebuilt around AI. The people who suffer most will be those whose tasks are automated before they have the experience to move up to the tasks that remain.

Is the $15.7 trillion AI economy real?

The $15.7 trillion figure is a projection, not a result, and it is nine years old. PwC published it in 2017 in a report called Sizing the Prize. It forecast that AI would make global GDP 14% higher in 2030, with China gaining 26% and North America 14.5%. Together, those two regions would capture almost 70% of the gain, about $10.7 trillion. PwC expected more than half of the boost to come from labour productivity, and the rest from consumer demand for better products.

The investment side of that bet has arrived. Stanford’s 2026 AI Index reports that US AI investment reached $285.88 billion in 2025, compared with $20.92 billion in Europe and $12.41 billion in China. The same report says generative AI reached 53% adoption within three years, faster than the internet or the personal computer.

The profit side is slower. McKinsey’s 2025 survey of 1,993 people in 105 countries found that 88% of organisations use AI in at least one business function, up from 78% a year earlier. Yet only 39% could attribute any EBIT impact to it, and most of those said the impact was under 5%. Only about 6% qualified as “high performers” earning 5% or more of EBIT from AI.

Official statistics are more sober still. The US Census Bureau’s business survey found that 19.8% of US businesses were using AI in the period ending 3 May 2026, rising to 37% among firms with at least 250 employees. The gap between the 88% in consulting surveys and the 19.8% in government data tells you something useful: big companies are experimenting, while most small businesses have barely started.

So, is the prize real? The spending is real, the productivity is starting to show, and the full $15.7 trillion remains unproven. I would treat it as a direction of travel rather than a figure to quote as fact.

Who is AI hurting first?

Early-career workers in automatable, desk-based jobs are feeling it first. The clearest evidence comes from the Stanford study by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, published in August 2025. Using payroll records from the largest payroll software provider in the US, they found a 13% relative decline in employment for workers aged 22 to 25 in the most AI-exposed occupations. Older workers in the same occupations stayed stable or grew. The losses were concentrated where AI tends to automate tasks rather than assist people with them.

PwC’s 2026 AI Jobs Barometer points the same way. It reports that early-career job postings have flatlined in highly AI-exposed sectors, and that the most exposed junior roles are seven times more likely than the least exposed to ask for senior skills such as leadership. In plain language, the first rung of the ladder is getting higher.

Layoff announcements add another signal, though a noisier one. According to Challenger, Gray & Christmas, AI was the leading stated reason for US job cuts in 2026 through August, at about 22% of all announced cuts. Total announced cuts were actually down 41% from the same period in 2025. Companies may also be citing AI to explain cuts they would have made anyway, so I read this as a trend in corporate language as much as in technology.

This is the part of the story that the books I read for the first version mostly missed. Martin Ford, Darrell West and others warned about displacement in general. Few predicted that the first visible damage would fall on graduates rather than on factory workers.

Which jobs are shrinking and which are growing?

Clerical and routine creative jobs are shrinking fastest, while care, education, frontline work and technical specialist roles are growing. The World Economic Forum’s Future of Jobs Report 2025, based on a survey of more than 1,000 employers across 55 economies, expects 170 million new jobs and 92 million displaced jobs by 2030, a net gain of 78 million. It names cashiers, administrative assistants and graphic designers among the fastest-declining roles.

Declining roles (WEF 2025)	Growing roles (WEF 2025)
Cashiers and ticket clerks	Farmworkers, delivery drivers and construction workers
Administrative assistants and executive secretaries	Nursing professionals and other care roles
Graphic designers	Secondary school teachers
Routine data and clerical work	AI, big data and cybersecurity specialists
Routine customer service scripts	Renewable energy and environmental engineers

The growing list surprises people. Many of the fastest-growing jobs are physical, local or relational: harvesting food, delivering goods, nursing patients, teaching teenagers. AI is good at text and images. It is still poor at lifting a patient, calming a classroom or fixing a roof.

The WEF also expects 22% of jobs to be disrupted and nearly 40% of on-the-job skills to change by 2030. Of every 100 workers, it estimates 59 will need training, and 11 of them are unlikely to get it. Those 11 are the people public policy should worry about most.

Which jobs might disappear because of AI?

Clerical jobs built on routine data handling are the most likely to shrink sharply, and some may all but disappear in their current form. The World Economic Forum’s employer survey projects these as the fastest-declining jobs between 2025 and 2030, by percentage:

Job	Projected change, 2025 to 2030
Postal service clerical workers	34% fewer
Bank tellers and related clerical workers	31% fewer
Data entry clerks	26% fewer
Cashiers and ticket clerks	20% fewer
Administrative assistants and executive secretaries	20% fewer
Printing and related trades workers	20% fewer
Accounting, bookkeeping and payroll clerks	18% fewer
Material-recording and stock-keeping clerks	16% fewer
Door-to-door sales workers, news and street vendors	14% fewer

Not all of this decline is caused by AI. Digital payments, online shopping and self-service machines play a part too. But generative AI speeds it up, because these jobs are made mostly of tasks that software can now read, sort, write and check.

A second list comes from Microsoft Research, which studied 200,000 anonymised conversations with Bing Copilot to see which occupations AI can already help with most. Its highest “AI applicability” scores went to these jobs, among others: interpreters and translators, historians, passenger attendants, sales representatives of services, writers and authors, customer service representatives, telephone operators, ticket agents, broadcast announcers, telemarketers, news analysts and journalists, technical writers, and proofreaders.

Read that list carefully. A high score means AI overlaps with much of the work, not that the job is gone. Translators and writers are already seeing fewer routine commissions and more editing work. Customer service, telemarketing and proofreading are the roles where I would expect the steepest cuts, because the task is narrow and the output is easy to check.

What new jobs is AI creating?

AI is creating demand for people who build, feed, check and govern it, and for digital specialists more broadly. The WEF’s list of the fastest-growing jobs between 2025 and 2030 is led by technical roles:

Job	Projected growth, 2025 to 2030
Big data specialists	113%
Fintech engineers	93%
AI and machine learning specialists	82%
Software and applications developers	57%
Security management specialists	53%
Data warehousing specialists	49%
Autonomous and electric vehicle specialists	48%
UI and UX designers	48%
Light truck and delivery drivers	46%
Internet of Things specialists	42%
Data analysts and scientists	41%

Job-site data shows the same pattern. LinkedIn’s 2026 Jobs on the Rise list for the United States put AI engineer first, with AI consultant and data annotator also among the fastest-growing roles. In Australia, the same list featured AI engineer at the top and a surge in director of artificial intelligence roles.

Some of the most interesting new jobs are the ones Paul Daugherty and H. James Wilson predicted in Human + Machine. They described three kinds of human work that AI needs:

• Trainers, who teach AI systems what good output looks like. Data annotators and AI evaluators are this job today.
• Explainers, who translate what an AI system did, and why, for managers, customers and regulators. AI consultants and AI governance staff fill this role.
• Sustainers, who keep AI working safely and fairly over time. AI risk, compliance and security specialists belong here.

Not every new job is technical. Because the WEF counts all trends, not only AI, its largest-growing jobs by sheer numbers are farmworkers, with 34 million more jobs expected by 2030, followed by delivery drivers, software developers, building construction workers and shop salespersons.

Which jobs will AI never replace?

No job is guaranteed forever, but jobs that need skilled hands in unpredictable places, physical care, or human trust and accountability are the hardest for AI to replace. Microsoft Research’s study gave these occupations some of the lowest AI applicability scores of all, close to zero:

• Skilled physical and outdoor work: dredge operators, bridge and lock tenders, pile driver operators, rail-track laying and maintenance operators, logging equipment operators, motorboat operators, roustabouts in oil and gas, roofers and their helpers, paving and surfacing equipment operators, and floor sanders and finishers.
• Plant and process operation: water treatment plant operators, gas compressor and pumping station operators, foundry mould and coremakers, and tyre builders.
• Hands-on care and health: orderlies, surgical assistants, ophthalmic medical technicians, massage therapists, and maids and housekeeping cleaners.

To that list I would add the human-centred jobs that the WEF expects to keep growing: nurses, social workers, counsellors and secondary school teachers. AI can draft a lesson plan or a care note. It cannot sit with a frightened patient, earn a teenager’s trust or take legal responsibility for a decision.

There is an honest caveat. A low score today measures what people ask chatbots to do, not what robots might do in twenty years. Physical jobs are protected mainly by the slow progress of robotics, and care jobs by the fact that people want a person. The safest strategy is not to pick a “safe” job and stop learning. It is to combine work that needs your presence with the ability to use AI for the paperwork around it.

Do AI skills actually raise your pay?

Yes, for now. PwC’s 2025 Global AI Jobs Barometer analysed close to a billion job ads and found that roles requiring AI skills paid an average 56% more than similar roles without them in 2024, up from 25% the year before. Productivity growth in the most AI-exposed industries rose to 27% over 2018 to 2024, compared with 9% in the least exposed.

That report also contains a detail that cuts against the doom story. Jobs in more AI-exposed occupations still grew, by 38% between 2019 and 2024, though less than the 65% growth in less-exposed occupations. Exposure slowed hiring. It did not reverse it.

The 2026 edition adds a useful distinction. PwC splits jobs into two groups. In “professionalised” jobs, AI raises the level of human skill needed. In “democratised” jobs, AI lets non-experts do the work. Professionalised jobs are growing twice as fast and have seen 42% faster wage growth since 2021. New tasks added to AI-exposed roles are 2.5 times more likely to rely on empathy, judgement or creativity.

That is the most practical finding in this whole article. The pay is going to people who use AI to do harder work, not to people who let AI do their existing work for them.

How are companies really using AI at work?

Most companies are still experimenting, and only a minority have redesigned their work around AI. McKinsey found that about one-third of organisations had begun scaling AI in 2025. On AI agents, the software that can carry out multi-step tasks, 23% said they were scaling an agent somewhere in the business and 39% were experimenting. In any single business function, no more than 10% reported scaling agents.

Leaders expect much more. Microsoft’s 2025 Work Trend Index found that 82% of leaders expect to use “digital labour” to expand their workforce within 12 to 18 months, and 41% expect their teams to be training AI agents within five years.

The way people use AI also matters. When Anthropic studied millions of anonymised conversations with its Claude assistant, it found that 57% of tasks were augmentation, where AI works with the user, and 43% were automation. Software development dominated, at 37.2% of queries. About 36% of occupations showed AI use in at least a quarter of their tasks.

Workforce plans are mixed rather than brutal. In McKinsey’s survey, 32% of respondents expected their company’s headcount to fall by 3% or more over the next year, 43% expected no change and 13% expected an increase. The WEF’s employer survey shows the same tension: 77% plan to upskill their workers, while 41% plan to reduce staff where AI automates tasks.

How do workers feel about AI?

More worried than hopeful. A Pew Research Center survey of 5,273 employed US adults found that 52% feel worried about AI’s future use in the workplace and 36% feel hopeful. A third said they felt overwhelmed. Only 16% said at least some of their work was done with AI, and 32% expected AI to mean fewer job opportunities for them.

I understand the worry. When I read layoff memos that credit “efficiency gains from AI,” I notice how rarely they explain which tasks were automated. Fear grows in that gap. Companies that explain plainly what AI will and will not do in each role give their staff something better than fear to work with.

What does AI mean for Bangladesh and other developing economies?

Developing economies face less immediate exposure but a deeper long-term risk. The IMF estimates AI exposure at 40% of jobs in emerging markets and 26% in low-income countries, against 60% in advanced economies. Lower exposure sounds like protection. It also means fewer workers in a position to gain from AI, and the IMF warns this could widen the gap between countries.

As someone who writes from Bangladesh, I cannot ignore the garment question. A study by the government’s a2i programme and the ILO, reported in 2018, projected that automation and digitalisation could threaten 5.38 million jobs across five Bangladeshi sectors by 2041, including about 2.7 million garment jobs. That study predates generative AI and focused on factory automation, so its numbers should be read as a warning rather than a forecast. Still, the logic holds. Kai-Fu Lee argued in AI Superpowers that countries which grew on cheap labour could lose that advantage as machines become cheaper than people.

For Bangladesh, the next decade’s question is whether its young graduates can move into the AI-assisted service work that is replacing the old ladder.

Are governments ready for AI at work?

Not yet, and some are slowing down. The European Union’s AI Act was meant to impose its strictest rules on high-risk systems, including AI used in hiring and worker management, from August 2026. Amendments finalised in 2026 moved the main high-risk obligations under Annex III to 2 December 2027, with Annex I systems following on 2 August 2028.

Public trust is low in the country building most of the technology. Stanford’s 2026 AI Index found that only 31% of US respondents trust their government to regulate AI, the lowest of the countries surveyed, compared with 91% in Singapore. The same report counted a record 362 reported AI incidents in 2025.

The policy ideas from the books still stand: portable benefits for gig workers, lifelong learning accounts, and stronger social insurance for people between jobs. Darrell West argued in The Future of Work that education must become a lifelong process rather than a one-time investment. The WEF’s finding that 11 in every 100 workers may never get the training they need is the strongest case yet for that idea.

What did the AI books get right, and what did they get wrong?

The books were right about the direction of change and wrong about its order. Here is my scorecard after re-reading them against the 2026 data.

Right: humans and machines work best together. Paul Daugherty and H. James Wilson’s Human + Machine described a “missing middle” where people train, explain and supervise AI. PwC’s finding that professionalised jobs are growing fastest is that idea showing up in wage data.

Right: the control problem matters. Stuart Russell’s Human Compatible and Brian Christian’s The Alignment Problem argued that AI pursuing the wrong objective is the deeper risk. At work, that looks like biased hiring tools and confident but wrong answers. McKinsey found that 51% of AI-using organisations had experienced at least one negative consequence, with inaccuracy the most common.

Half right: mass displacement. The 2013 estimate that 47% of US jobs were at risk and McKinsey’s 2017 estimate that 375 million people might need to switch occupations by 2030 described exposure, not destruction. The data so far shows targeted damage, not collapse.

Wrong: blue-collar jobs go first. Most automation writing assumed robots would empty factories before software touched offices. Generative AI reversed that. Writing, coding, design and customer support moved first.

Still unproven: the singularity. Ray Kurzweil’s prediction in The Singularity Is Nearer that human and machine intelligence will merge by 2045 remains a philosophical claim. Agents are improving fast. On one computer-use benchmark, accuracy rose from about 5% in 2024 to 66.3% in 2025, still below the 72.35% human baseline. That is impressive progress, and it is not a merger of minds.

If you want the book-by-book version of these debates, I have covered several of them separately, including whether AI could really wipe out humanity, Garry Kasparov’s argument in Deep Thinking, and a reading list of 20 books about AI and attention.

How can you prepare for an AI-shaped career?

Move toward the parts of your job that need judgement, accountability and trust, and learn to direct AI on the rest. These five steps follow directly from the evidence above.

Audit your own tasks. List what you do in a typical week and mark which tasks are routine text, data or image work. Those are the tasks AI will take first. Plan to spend less time on them within a year.

Use AI daily on real work, not tutorials. The 56% wage premium goes to people who can show results. Keep a record of time saved or quality improved, and put it in your CV with numbers.

Build the skills AI adds to your job. PwC found new tasks in exposed roles are 2.5 times more likely to need empathy, judgement or creativity. Volunteer for client conversations, reviews and decisions, not only production work.

If you are early in your career, get supervised experience fast. Entry-level tasks are the ones disappearing. Internships, apprenticeships and small real projects matter more than another certificate.

Keep one foot in work AI cannot easily reach. Care, teaching, skilled trades and local services are on the WEF’s growth list. A second skill in one of these areas is insurance, not a step down.

Students and parents asking about schools should also read my piece on whether AI makes students dumber or smarter, because the answer depends heavily on how AI is used in class.

Frequently asked questions about AI and the future of work

The bottom line on AI and the future of work

When I wrote the first version of this article, I thought the central question was how many jobs AI would destroy. The evidence since then has changed my mind. The better question is who gets to climb. AI is not emptying offices. It is removing the bottom rungs of many careers while raising the pay of people who can direct it well. That is a solvable problem, but only if employers keep hiring juniors on purpose, schools teach judgement alongside tools, and governments fund the training that 11 in every 100 workers may otherwise never receive.

The $15.7 trillion prize may or may not arrive on schedule. The people who will share in it are already visible: those who treat AI as a colleague to supervise rather than a rival to fear.

Sources

• World Economic Forum, Future of Jobs Report 2025 (January 2025)
• Kristalina Georgieva, IMF Blog (January 2024)
• International Labour Organization, Generative AI and Jobs: A 2025 Update (May 2025)
• Goldman Sachs, How Will AI Affect the Global Workforce? (August 2025)
• Brynjolfsson, Chandar and Chen, Canaries in the Coal Mine? (Stanford, August 2025)
• McKinsey, The State of AI in 2025 (November 2025)
• Stanford HAI, AI Index Report 2026 (April 2026)
• US Census Bureau, Business Trends and Outlook Survey (May 2026)
• PwC, 2025 Global AI Jobs Barometer (June 2025)
• PwC, 2026 AI Jobs Barometer (June 2026)
• Euronews on the WEF fastest-growing and fastest-declining jobs (February 2025)
• World Economic Forum, Jobs of the Future (January 2025)
• Tomlinson et al., Microsoft Research, Working with AI (2025)
• Let’s Data Science on LinkedIn Jobs on the Rise 2026 (January 2026)
• Startup Daily on LinkedIn Jobs on the Rise 2026, Australia (January 2026)
• World Economic Forum on PwC’s Sizing the Prize (June 2017)
• Challenger, Gray & Christmas, Job Cut Report (August 2026)
• Pew Research Center, US Workers and AI (February 2025)
• Microsoft, 2025 Work Trend Index (April 2025)
• Anthropic, The Anthropic Economic Index (February 2025)
• Orrick, EU AI Act Digital Omnibus Update (July 2026)
• The Daily Star on the a2i and ILO automation study (December 2018)
• Frey and Osborne, The Future of Employment (2013)
• MIT Technology Review on McKinsey’s 2017 automation study
• Books: Klaus Schwab, The Fourth Industrial Revolution; Darrell M. West, The Future of Work; Nick Bostrom, Superintelligence; Stuart Russell, Human Compatible; Brian Christian, The Alignment Problem; Kai-Fu Lee, AI Superpowers; Paul Daugherty and H. James Wilson, Human + Machine; Ray Kurzweil, The Singularity Is Nearer; Martin Ford, Rise of the Robots.