Will AI Replace Jobs? What the Data Actually Shows in 2026
The anxiety is real. In a Pew Research survey of 5,273 US workers published in February 2025, 52% said they feel worried about how AI may be used in the workplace in the future, 36% felt hopeful, and 32% expected AI to mean fewer job opportunities for them. That tension between dramatic forecasts and modest measured effects defines the current moment. Predictions of mass displacement have dominated headlines, yet macroeconomic data tells a quieter story. This article separates what AI could do to jobs from what has actually happened so far. It covers the main forecasts and what they measure, the hard evidence from payroll and labor-force data, which jobs are shrinking and which are growing, why young workers feel it first, the layoffs companies blame on AI, what history suggests, and our own analysis of 168 million US jobs sorted by AI exposure.
The short answer
The data from 2023 through mid-2026 paints a nuanced picture of labor market change. Here are the evidence-based takeaways:
- Exposure is not replacement. While many jobs have tasks that AI can perform, full job elimination is rare. The ILO estimates that only 0.4% to 5.5% of global employment is highly automatable, depending on income levels.
- Entry-level workers are feeling the shift first. A Stanford Digital Economy Lab study (revised August 2026, with payroll data through June 2026) puts employment of 22–25-year-olds in highly exposed occupations about 19% below where it would otherwise be, driven by fewer hires rather than layoffs.
- The big picture is still calm. The Yale Budget Lab finds no clear AI footprint in US labor market data, and a Danish study rules out effects larger than 2% on earnings or hours.
- AI mostly changes tasks, not headcount, so far. Productivity studies show large gains for some tasks, and most firms that adopt AI retrain staff rather than cut them, according to New York Fed surveys.
- Hands-on jobs are the least exposed. Construction, care, cleaning and the trades score near zero on usage-based measures of AI exposure, and many of them are projected to grow.
- Clerical and routine office work is shrinking. The US Bureau of Labor Statistics expects office and administrative support to lose 752,100 jobs by 2035, partly because of AI-powered automation.
What the forecasts say
To understand the fear, we must look at what the major institutions are predicting. However, a crucial distinction must be made: exposure estimates measure the potential for tasks to be affected, not the number of people who will lose their jobs.
The foundational estimate came in March 2023 from Tyna Eloundou and colleagues, in a paper written with OpenAI researchers. Their paper, “GPTs are GPTs,” estimated that about 80% of the U.S. workforce could have at least 10% of their work tasks affected by LLMs. Roughly 19% of workers could see at least 50% of their tasks affected. Notably, higher-income jobs were potentially more exposed than lower-income ones. The study was published in Science in June 2024.
Goldman Sachs followed with a widely cited March 2023 report estimating that generative AI could expose the equivalent of 300 million full-time jobs to automation globally. They projected that this could add 7% to global GDP. However, in August 2025, the same research team released a more tempered baseline forecast: 6–7% of the U.S. workforce could be displaced if AI is widely adopted, with a range of 3–14%. They noted that unemployment would likely run only 0.5 percentage points above trend during the transition period.
The International Monetary Fund (IMF) estimated in January 2024 that almost 40% of global employment is exposed to AI. In advanced economies, that figure rises to 60%, but roughly half of those exposed jobs may actually benefit from productivity gains rather than being eliminated. The International Labour Organization (ILO) provided a more granular view in 2023 and refined it in 2025. Their 2025 working paper found that 25% of global employment is in occupations with some GenAI exposure. In high-income countries, that figure is 34%. Crucially, the ILO concluded that the main effect is “transformation, not replacement,” with clerical work being the most exposed sector.
McKinsey Global Institute estimated in July 2023 that activities accounting for up to 30% of U.S. work hours could be automated by 2030. By November 2025, they updated this to say that 57% of U.S. work hours are technically automatable with today’s technology and redesigned workflows. However, they emphasized that this is not a job-loss forecast; more than 70% of employer-sought skills are used in both automatable and non-automatable work.
The World Economic Forum’s Future of Jobs Report 2025 (released January 2025) projected that over 2025–2030, 170 million jobs would be created while 92 million would be displaced, resulting in a net gain of 78 million jobs. This suggests a high rate of structural churn rather than net job loss.
The table below summarizes the headline estimates versus what they actually measure.
| Source | Year | Headline estimate | What it actually measures |
|---|---|---|---|
| Eloundou et al. | 2023 | 80% of workforce exposed | Percentage of tasks affected by LLMs |
| Goldman Sachs | 2023 | 300 million jobs exposed | Global equivalent full-time jobs with affected tasks |
| IMF | 2024 | 40% global employment exposed | Share of jobs where tasks could be augmented or automated |
| ILO | 2025 | 1 in 4 jobs exposed | Occupations with some GenAI exposure, primarily clerical |
| McKinsey | 2025 | 57% of US hours automatable | Technical automatability of work hours, not job loss |
| WEF | 2025 | 92 million displaced, 170M created | Net change in job volume over 5 years |
What has actually happened so far
If the forecasts are so high, why doesn’t the labor market look like it is collapsing? The answer lies in the difference between what can happen and what has happened.
The Yale Budget Lab, in a report released October 2025, found that the broader labor market had not experienced a discernible disruption since ChatGPT’s release 33 months prior. Their October 2026 tracker update confirmed that the occupational mix is not yet changing in ways that clearly align with the introduction of AI. Exposure, automation, and augmentation measures show no connection to changes in employment or unemployment at the macro level.
Surveys of firms point the same way. Not every Fed analysis is so reassuring: a St. Louis Fed note from August 2025 found a correlation of 0.57 between how much occupations adopted AI and how much their unemployment rose from 2022 to 2025, the strongest link in computer and math jobs, while stressing that correlation is not proof. A New York Fed survey in September 2025 found that while AI use in service firms rose to 40%, AI-related layoffs remained uncommon at just 1%. The primary response to AI was retraining, not reduction.
So why do we hear about layoffs? The data from Challenger, Gray & Christmas shows that in 2025, companies cited AI for 54,836 of 1.2 million job cuts. By September 2026, AI was cited as the reason for 3,961 cuts in a single month, totaling 120,136 year-to-date. While this sounds significant, it amounts to only 0.07% of the 170 million U.S. jobs. Furthermore, Challenger itself warns of “AI-washing,” where companies cite AI to justify cuts that might otherwise be attributed to broader efficiency drives or cost-cutting. This is supported by New York’s WARN act data, where zero companies checked the AI box in their first year of reporting.
The most concrete evidence of actual labor market impact comes from the Stanford Digital Economy Lab. Their August 2025 paper found a 13% relative employment decline for 22–25-year-olds in the most AI-exposed occupations. By their August 2026 revision, covering data through June 2026, that gap had widened to 19% below the counterfactual. However, this was driven by reduced hiring, not separations. Experienced workers showed no gap. The decline was concentrated where AI automates tasks, while roles where AI complements human work remained flat or grew.
This aligns with the work of Humlum and Vestergaard, who studied the Danish labor market. Their NBER working paper, revised in March 2026, found “precise zeros” on earnings and recorded hours effects two years after ChatGPT’s release. They concluded that firms are absorbing AI by reorganizing tasks rather than cutting headcounts.
In summary, the macro data looks calm because AI is currently an augmentative tool that changes how work is done, not a wholesale replacement of who does it. The dramatic stories we see are often at the margins—specifically affecting young workers entering the market or in highly automatable roles like customer service and data entry.
Which jobs AI is most likely to replace
When we look at which jobs AI is most likely to replace, the data points to a clear pattern: routine, repetitive, and text-heavy tasks are the first to go. Government projections now build this in.
The US Bureau of Labor Statistics (BLS), in its Employment Projections 2025–35 released in August 2026, identifies office and administrative support as the fastest-declining group. This sector is projected to shrink by 4.0%, losing 752,100 jobs over the next decade, partly due to AI-powered automation. Specific roles within this group are facing steep declines: word processors and typists are expected to fall by 34.4%, data entry keyers by 25.5%, and telemarketers by 21.4%.
Similarly, customer service representatives, a role with 2.67 million jobs in 2025, are projected to decline by 5% (a loss of 141,800 positions) because tasks “continue to be automated.” Computer programmers, a role with 110,800 jobs in 2025, face a projected 7% decline as companies use AI “to automate repetitive programming tasks.”
The World Economic Forum’s Future of Jobs Report 2025 supports this trajectory. It lists data entry clerks, cashiers, administrative assistants, and accounting clerks among the fastest-declining roles.
Exposure estimates from leading research firms align with these projections. The original “GPTs are GPTs” study by Eloundou et al. (March 2023) found that about 19% of US workers could see at least 50% of their tasks affected by LLMs, with computer programmers scoring 0.95 on a 0–1 scale. A 2025 Microsoft Research paper, analyzing 200,000 Bing Copilot conversations, placed Customer Service Reps at 0.408 and Telemarketers at 0.404 on its AI applicability score.
It is important to distinguish between “exposure” and actual job loss. Exposure measures how much of a job’s tasks could be done by AI. It does not mean the job disappears overnight. However, when exposure meets economic pressure, the result is often reduced hiring or slower growth in those roles.
| Job | 2025 US employment | BLS projection 2025-35 | Why (from BLS/WEF) |
|---|---|---|---|
| Office clerks, general | 2,600,000 | −6.0% | Office and admin support shrinking, partly due to AI automation |
| Secretaries and administrative assistants | 1,890,000 | −6.0% | Same group |
| Data entry keyers | 131,800 | −25.5% | Among the fastest-declining occupations |
| Telemarketers | 60,900 | −21.4% | Among the fastest-declining occupations |
| Customer service reps | 2,666,000 | −5% | Tasks “continue to be automated” |
| Computer programmers | 110,800 | −7% | AI automates repetitive coding tasks |
Our analysis: 168 million US jobs, sorted by AI exposure
Forecasts and anecdotes are easy to find; a simple cross-check is not, so we built one. On 18 October 2026 we joined three public datasets by occupation code: the US Bureau of Labor Statistics’ 2025 employment and 2025–35 projections for 831 occupations, Microsoft Research’s “AI applicability” score for 785 occupations (based on 200,000 real Copilot conversations), and Anthropic’s “observed exposure” score for 756 occupations (based on what people actually use Claude for at work). We then sorted occupations by each score and split US employment into five equal groups, from most to least exposed, and compared what BLS expects for each group over the next decade.
| Exposure group (Microsoft score) | US jobs, 2025 | BLS projected growth, 2025–35 | Median wage |
|---|---|---|---|
| Most exposed fifth | 34.1 million | +1.1% | $64,000 |
| Second | 33.7 million | +1.5% | $68,800 |
| Middle | 33.1 million | +5.4% | $73,100 |
| Fourth | 33.6 million | +4.2% | $68,300 |
| Least exposed fifth | 33.4 million | +5.5% | $49,200 |
All US occupations together are projected to grow 3.5%. The Anthropic measure tells the same story: the most exposed fifth (32.1 million jobs) is projected to grow just 0.7%, against 2.8% to 5.3% for the other groups, and 61 million jobs sit in occupations where Anthropic sees no Claude usage at all.
Four things stand out.
The most exposed jobs are expected to stall, not vanish. Growth of about 1% over ten years is weak, but it is still growth. The pain is concentrated in specific large occupations: customer service representatives (2.67 million jobs, −5.3%), general office clerks (2.6 million, −6.0%), secretaries and administrative assistants (1.89 million, −6.0%), data entry keyers (−25.5%) and telemarketers (−21.4%).
High exposure is not a death sentence. Several highly exposed occupations are projected to grow: management analysts (+10.1%), market research analysts (+7.0%), computer systems analysts (+7.9%), financial analysts (+7.2%) and database architects (+9.4%). In these jobs AI takes over tasks while demand for the work keeps rising.
The safest jobs are physical, and many pay less. The least exposed fifth has the lowest median wage, $49,200, and includes maids and housekeepers, dishwashers, nursing assistants, construction laborers, forklift operators and landscapers. They are protected less by their pay or prestige than by the fact that software cannot lift, clean or drive. The most exposed fifth earns more, but the best-paid group is the middle, where people combine knowledge work with judgment, physical presence or responsibility.
The two measures disagree on details. Microsoft’s score ranks interpreters, writers and sales representatives highest; Anthropic’s ranks computer programmers (0.745) and customer service representatives (0.701) highest. Both find almost no exposure in hands-on work.
How to read this: BLS projections are expert judgments made occupation by occupation, and they already assume some AI adoption, so this is a consistency check, not proof that AI causes the slower growth. Exposure measures what AI can help with, not whether a job disappears. Our figures cover 784 to 755 matched occupations and 154 to 168 million jobs, depending on the measure.
Jobs AI can’t easily replace
If you work with your hands, in unpredictable physical environments, or in direct human care, AI is unlikely to replace you anytime soon. The data on physical, in-person, and skilled-trade work shows scores near zero on usage-based measures.
The Microsoft Research applicability scores highlight this gap. Roles like Dredge Operators, Bridge and Lock Tenders, Floor Sanders, and Orderlies all score 0.0. Anthropic’s economic index found that about 30% of workers—cooks, bartenders, and lifeguards—have zero AI exposure because their tasks are highly physical and context-dependent.
The BLS projects strong growth in these sectors. Nurse practitioners are expected to grow by 41.0%. Solar PV installers will see a 36.5% increase. Wind turbine technicians are projected to grow by 29.5%. Home health and personal care aides will grow by 18.1%, from 4.68 million to 5.52 million jobs, the largest absolute gain of any occupation.
Why do these jobs resist automation? Robotics lags behind software. McKinsey notes that while digital agents can automate 44% of work hours, robots account for only 13%. Physical jobs require fine motor skills, endurance, and adaptability to unstructured environments that current robots cannot match at scale. The WEF’s largest absolute growth roles include farmworkers, delivery drivers, and construction workers, all relying on physical presence.
Care work is another stronghold. Nursing professionals, social workers, and personal care aides are expected to grow significantly. The WEF explicitly notes that care roles are among the largest growth areas. Human empathy, trust, and complex physical interaction are hard to automate.
New AI-created roles are also emerging. Indeed data shows that “forward-deployed engineers” rose from 643 listings in April 2025 to 5,330 in April 2026. Lightcast counted US postings for the role rising from under 200 in 2023 to more than 7,300 in January–September 2026 (both figures are as reported in press coverage). The WEF lists AI/ML specialists among the top three fastest-growing roles. These are specialist roles counted in thousands, not the millions of jobs in routine office work.
The key takeaway: AI replaces tasks, not always entire jobs. If your job is 80% routine data processing, it is vulnerable. If it is 80% physical work or hands-on care, it is far less exposed.
Young and entry-level workers feel it first
The most immediate impact of AI is not on layoffs, but on hiring. Young and entry-level workers are feeling the shockwaves first.
The Stanford Digital Economy Lab’s “Canaries in the Coal Mine” study provides the clearest evidence. The original August 2025 paper found a 13% relative employment decline for 22–25-year-olds in the most AI-exposed occupations. The August 2026 revision, using data through June 2026, shows this gap has widened: employment for this group is now about 19% below the counterfactual.
Crucially, this is not because companies are firing experienced workers. The gap comes from reduced hiring. Experienced workers are unaffected. The decline is concentrated where AI automates routine tasks, and it shows up in headcount, not base pay.
Goldman Sachs research supports this. Their August 2025 report noted that unemployment among 20–30-year-olds in tech-exposed occupations is up almost 3 percentage points since early 2025. Indeed data shows US software development postings have fallen to 78.1 (index Feb 1, 2020 = 100), down from a peak of 233.8 in early 2022. The index does not separate AI from the hiring correction that followed the 2021–22 tech boom.
The OECD is more cautious: in August 2026 it said current evidence does not show that AI is responsible for weaker youth prospects, and that the trend predates large language models. Anthropic’s economic index found that job-finding rates for 22–25-year-olds into exposed occupations dropped about 14% versus 2022.
The worry is a squeeze on the bottom rung: entry-level roles have traditionally been where people learn a profession. Workers may find it harder to get their first break, potentially slowing long-term career progression. The WEF and PwC warn that more than one in three young workers are in medium-to-high exposure occupations.
Nobody yet knows whether this is a lasting change or a slower start to careers. The Yale Budget Lab still sees no clear AI footprint in the overall occupational mix, and new kinds of entry-level work may appear as AI tools spread, as happened with earlier technologies.
Layoffs blamed on AI: what companies say
When employers cite artificial intelligence as the reason for cutting staff, the story is rarely as simple as “robots took the job.” A closer look at corporate filings, press releases, and executive statements reveals a mix of genuine workflow shifts, cost-cutting rebranding, and what researchers call “AI-washing”—naming AI to signal efficiency to investors.
The most cited example is Klarna. In February 2024, the fintech company announced that its AI customer service assistant handled 2.3 million chats in its first month, by Klarna’s own estimate the work of 700 full-time agents and a $40 million profit improvement. Its March 2025 SEC filing (F-1) shows full-time employees falling from 5,527 in 2022 to 3,422 in 2024. Revenue per employee rose from roughly $344,000 to $972,000. However, the company did not stick to its all-AI strategy. By May 2025, the CEO reversed course, hiring human agents again because AI-only service had given “lower quality”, noting that the headcount decline was mostly due to attrition and hiring freezes rather than direct replacement.
Other major firms have been more transparent about the nuance. IBM paused hiring for roughly 26,000 back-office roles in May 2023, with CEO Krishna estimating that “30% of that getting replaced by AI” over five years. By May 2025, however, Krishna clarified that while AI replaced about 200 HR roles, total employment had actually gone up. Salesforce CEO Marc Benioff said on a podcast in 2025 that support headcount went from 9,000 to about 5,000 because AI agents handle about half of conversations; the company described it as not backfilling roles and redeploying staff.
In contrast, some cuts were explicitly framed as structural, not AI-driven. Microsoft cut roughly 15,000 roles in 2025 and another 4,800 in July 2026, with leadership stating the cuts were not “being replaced by AI.” Atlassian also cut 1,600 roles in March 2026, with its CEO explicitly stating it was not an “AI replaces people” scenario. Oracle, however, provided a more direct link in its 10-K filing, stating that “The adoption and deployment of AI technologies across our operations have resulted” in workforce reductions, covering about 21,000 roles over 12 months.
Amazon CEO Andy Jassy issued a memo in June 2025 stating AI “will reduce our total corporate workforce.” This was followed by two rounds of cuts totaling roughly 30,000 corporate roles in late 2025 and early 2026, which leadership described as removing bureaucratic layers. Duolingo’s CEO sent an “AI-first” memo in April 2025, stating the company would stop using contractors for work AI could handle, though he later clarified they never laid off full-time staff. Block (formerly Square) cut 4,000 roles in February 2026, with Jack Dorsey’s letter crediting “intelligence tools.”
The ambiguity is widespread. Challenger, Gray & Christmas reported that AI was the top reason for job cuts in July 2026, accounting for 33% of 33,429 announced cuts. Yet by September 2026, AI cited as the reason dropped to the fifth spot, accounting for only 9% of cuts, though it remained the #1 reason year-to-date. In total, AI-cited cuts amounted to roughly 120,000 roles year-to-date, or about 0.07% of the 170 million US jobs. Challenger itself warns that naming AI can “win over investors,” a sentiment echoed by coverage of the Yale Budget Lab’s research on corporate “AI-washing.”
What history teaches
Predictions about technology displacing workers are old, and Goldman Sachs itself notes they have “a long history but a poor track record”. The most famous modern forecast came from Carl Frey and Michael Osborne in 2013, who estimated that 47% of US employment was at high risk of computerization. When the paper was published in 2017, that figure drew global attention. Yet, US employment actually grew by 10.9% between 2015 and 2025, reaching 170.3 million jobs. The economy did not shrink; it shifted.
A clearer lesson comes from the history of automation in banking. As ATMs spread (the US eventually had more than 400,000), you might have expected bank teller jobs to vanish. Instead, economist James Bessen found, the number of tellers per urban branch fell from 20 to 13 between 1988 and 2004, but the number of urban branches rose 43%, and teller employment did not fall. The ATM changed the task of the teller, allowing banks to open more branches and hire more people overall. Today, the Bureau of Labor Statistics (BLS) projects a 13% decline for tellers from 2025 to 2035, but this is driven by smartphones and online banking, not ATMs.
We also constantly forget that most jobs today did not exist a few decades ago. Research by Autor, Chin, Salomons, and Seegmiller (2024) found that roughly 60% of employment in 2018 was in job titles that did not exist in 1940. For professionals, that share is 74%. New technologies create new categories of work, such as data scientists, whom BLS now expects to grow 34.6% by 2035. That does not make the transition painless for the people whose tasks are automated, but it is why total employment has kept growing through each wave of automation.
Augmentation: the productivity evidence
The most robust evidence to date suggests AI is augmenting work more than it is automating it away. A landmark study by Brynjolfsson, Li, and Raymond (2025) analyzed 5,172 customer support agents and found that an AI assistant raised the number of issues resolved per hour by 14–15%. The gains were largest for less experienced workers: 34% for novices in the NBER working-paper version. This suggests AI acts as a force multiplier, helping novices perform at a higher level rather than replacing them outright.
Noy and Zhang (Science, 2023) found that 453 professionals using ChatGPT finished writing tasks 40% faster, with quality up 18%. Dell’Acqua and colleagues’ study of 758 BCG consultants described a “jagged frontier”: on tasks inside AI’s capabilities, consultants with AI completed 12.2% more tasks, 25.1% faster and at higher quality, but on a task outside it they were 19% less likely to reach the correct answer. Human judgment about when to trust the tool matters.
Anthropic’s Economic Index provides real-world usage data. In early 2025, 57% of AI usage was augmentation versus 43% automation. By September 2025, automation briefly surpassed augmentation at 49%, but by January 2026, augmentation rebounded to 52%. In Anthropic’s June 2026 survey of its users, 86% reported speed gains, 57% said AI raised the market value of their skills, and 10% thought losing their own job was likely. The data points to a future where AI handles the drudgery, allowing humans to focus on higher-value tasks, rather than a mass exodus of workers.
What governments are doing
Governments are starting to track AI’s impact on employment, though the data is still emerging. New York State added a checkbox to its WARN notice form in March 2025, asking if technological innovation contributed to layoffs. In the first year, about 160 companies filed, and none checked the box. Connecticut passed Public Act 26-15 in May 2026, requiring employers to disclose if AI caused layoffs starting October 2026. California’s SB 951, effective January 2027, will require similar disclosures, with summaries published by the EDD.
At the federal level, the “AI-Related Job Impacts Clarity Act” was introduced in November 2025, proposing quarterly reporting to the Department of Labor. It remains in committee. Meanwhile, OpenAI announced an OpenAI Jobs Platform and Certifications in September 2025, aiming to certify 10 million Americans by 2030, with Walmart as a launch partner, though the launch status remains unconfirmed. These measures aim to bring transparency to a market where companies are increasingly citing AI as a reason for restructuring.
How to protect your career
The data suggests that AI is reshaping tasks, not just eliminating jobs. Here is how to position yourself:
- Use AI in your current job: Don’t wait for it to be mandated. Learn how AI can automate the repetitive parts of your role, freeing you to focus on higher-value work. In the Noy and Zhang study, writing tasks took 40% less time.
- Move toward judgment and relationships: Jobs requiring physical presence (cooks, bartenders, lifeguards) or complex human judgment are less exposed. The Microsoft study shows near-zero AI applicability for roles like dredge operators or orderlies.
- Entry-level workers: Build proof of work: AI is great at generating generic content. Differentiate yourself by showing unique insight, context, and problem-solving. The Stanford “Canaries” study shows entry-level workers are most affected by reduced hiring, so visible competence is key.
- Managers: Retrain rather than cut: New York Fed surveys show that retraining is the primary response to AI adoption, not layoffs. Invest in upskilling your team to work with AI, rather than replacing them.
- Watch BLS projections: The BLS projects growth in healthcare (+41% for nurse practitioners), green energy (+36.5% for solar installers), and tech (+7.3% for computer/math). Declining roles are mostly administrative (−4% for office support). Align your skills with growing sectors.
Frequently asked questions
Will AI replace jobs? AI is replacing tasks, not entire jobs wholesale. While some roles like data entry and customer service are shrinking, most jobs are being augmented. The BLS projects overall employment growth, with AI primarily affecting administrative and routine cognitive roles.
What jobs will AI replace first? Jobs with high routine cognitive content are most exposed. According to the BLS, word processors, typists, and data entry keyers are declining rapidly. Customer service representatives and telemarketers are also seeing reduced demand as AI agents handle more conversations.
What jobs can’t AI replace? Jobs requiring physical dexterity and in-person interaction are least exposed. Cooks, bartenders, lifeguards, and construction workers score near zero on AI applicability indices. Roles requiring complex human judgment, empathy, and physical presence remain safe for the foreseeable future.
How many jobs has AI replaced so far? AI-cited layoffs have totaled roughly 120,000 year-to-date in 2026, representing about 0.07% of US jobs. While this sounds significant, it is a small fraction of the total workforce. Most AI impact so far has been in reduced hiring for entry-level roles rather than mass firings.
Is AI the reason it is hard for graduates to find jobs? Not entirely. The Yale Budget Lab and OECD note that weaker youth prospects predate generative AI. However, the Stanford “Canaries” study finds employment of 22–25-year-olds in the most AI-exposed occupations about 19% below where it would otherwise be, largely because of fewer hires. AI is a factor, but not the sole cause.
Will AI take software developer jobs? AI is automating repetitive programming tasks, leading the BLS to project a 7% decline in programmer jobs from 2025 to 2035. However, BLS projects data scientists to grow 34.6% and software developers about 10%, and Indeed’s software development postings index has recovered from its May 2025 low. Developers who adapt to use AI tools are becoming more productive, not obsolete.
Should I worry about my job? Worry is natural, but the data suggests adaptation is more effective than fear. Most jobs are being augmented, not eliminated. By learning to use AI tools and focusing on unique human skills like judgment and creativity, you can future-proof your career. The goal is to work with AI, not against it.