JOBS THAT WILL BE NEGATIVELY IMPACTED BY ARTIFICIAL INTELLIGENCE IN 2027

Curated by Artificial Intelligence

JOBS THAT WILL BE NEGATIVELY IMPACTED BY AI IN 2027 | PHOTO BY SALVATORE DE LELLIS

[JOBS IMPACTED BY AI ARTIFICIAL INTELLIGENCE] The best-supported answer is not that Artificial Intelligence will erase a single, predetermined list of occupations in 2027. It is that routine, digitally mediated jobs - especially clerical and transaction-processing work face the clearest risk of negative impact, while several customer-service and entry-level professional roles face a more conditional risk of weaker hiring and redesigned work. For this article, a job is negatively impacted when AI contributes to reduced hiring, wages, hours, autonomy, or access to employment opportunities. That definition deliberately includes deterioration in job quality and entry routes, not only layoffs.

Task exposure is not job elimination. Exposure means that AI is capable of performing or materially assisting some tasks in an occupation. Negative employment outcomes depend on whether firms actually deploy it, redesign the workflow, and need fewer paid labor hours after accounting for new demand and new human tasks.

The International Labor Organization (ILO) estimates that one in four workers globally is in an occupation with some generative-AI exposure, but only 3.3% of global employment is in its highest exposure gradient. Its central conclusion is that job transformation is more likely than full replacement, because most occupations retain tasks requiring human input. The International Monetary Fund (IMF) likewise estimates that nearly 40% of global employment is exposed but explicitly treats exposure as potentially complementary as well as substitutive. These are different measures, not competing forecasts of jobs lost.

THE 2027 JUDGMENT: RISK IS CONCENTRATED, NOT UNIVERSAL

The ranking below is a near-term risk assessment, not an assertion that these occupations will disappear during calendar year 2027. “Confidence” refers to confidence that the group will encounter adverse pressure on hiring or job quality by then; it does not claim confidence that AI alone will cause a net employment decline.

The strongest category is therefore clerical work, not “all white-collar work.” ILO’s 2025 index identifies data-entry clerks, typists, accounting and bookkeeping clerks, general office clerks, and administrative secretaries among the highest-exposure clerical roles. BLS baseline projections independently show steep 2025–35 declines for word processors and typists (34.4%), data-entry keyers (25.5%), and order clerks (17.5%). Those U.S. projections do not attribute the declines to AI, but they strengthen the case that these roles have little margin for another labor-saving shock.

Finance operations belong near the top of the list for the same reason: the evidence is about routine records and transactions, not finance as a whole. BLS projects a 6% decline for bookkeeping, accounting, and auditing clerks and a 4% decline for the broader financial-clerks group over 2025–35. It would be an overreach to turn this into a forecast that all accountants, analysts, advisers, or finance professionals will lose work.

WHY 2027 IS A NEAR-TERM INFLECTION POINT—NOT A FORECAST DEADLINE

No supplied source produces an occupation-by-occupation forecast specifically for 2027. The case for treating it as an inflection point is instead a convergence of adoption intentions, early labor-market signals, and a 2025–30 adjustment horizon.

First, WEF reports that 86% of surveyed employers expect AI and information-processing technologies to transform their businesses by 2030. Forty-one percent say they plan workforce reductions where AI automates particular tasks, but 77% plan to upskill workers and almost half expect to redeploy workers from exposed roles. Thus, 2027 sits early in a period when organizations may be moving from experimentation toward workflow redesign, not necessarily from employment to mass layoffs.

Second, the early empirical record suggests hiring may adjust before measured unemployment. Stanford researchers using administrative payroll data through June 2026 found no economy-wide displacement. They did find employment for 22–25-year-olds in AI-exposed occupations was 19% below the level implied by their less-exposed peers, with the gap driven primarily by reduced hiring rather than separations and concentrated where observed use was substitutive. The authors characterize this as an early descriptive indicator rather than proof that AI alone caused the change. Anthropic similarly found no systematic increase in unemployment among highly exposed workers, while estimating a barely statistically significant 14% relative decline in young workers’ job-finding rate into high-exposure occupations.

Third, a World Bank working paper examining 285 million U.S. online job postings found that, holding theoretical exposure comparable, occupations with greater AI-substitution vulnerability had 12% fewer postings than lower-vulnerability occupations. The posting gap rose from 6% in the first year after ChatGPT’s release to 18% by the third year and was especially large in administrative support and professional services. Online postings are not hires or employment, but the finding makes reduced entry hiring a plausible near-term channel of negative impact.

This is why 2027 should be read as a stress test for hiring pipelines and job design. It is too soon to infer broad, economy-wide redundancy. It is not too soon for employers, workers, and policymakers to watch vacancies, entry cohorts, hours, task allocation, and wage progression in the occupations above.

JOBS LIKELY TO BE PRESSURED, BUT NOT ERASED

Several visible occupations merit close attention without being placed in the top elimination-risk tier. Customer service is the clearest case. Automation may reduce the volume of routine contacts, but human agents remain valuable for escalation, empathy, judgment, and exceptions. A field study of 5,179 customer-support agents found access to an AI tool raised resolutions per hour by 14% on average and increased retention in that setting. This is evidence that the same technology can augment workers rather than simply replace them.

Writing, translation, and design also fit the “pressured, not erased” category. WEF includes graphic designers among fast-declining roles in its employer-expectation list, and ILO finds translators in its third exposure gradient. Yet the available U.S. occupational projections show no net decline for writers and authors and 2% projected growth for interpreters and translators. The defensible 2027 claim is narrower: AI may alter assignments, rates, production speed, and the viability of entry-level portfolio-building work. It is not defensible to say these occupations will vanish.

The same restraint applies to programming. BLS projects computer programmers to decline 7% between 2025 and 2035, while the occupation remains distinct from software developers and the wider technology workforce. The evidence supports concern about standardized coding tasks and junior hiring, not a claim that all software work is contracting.

WHAT COULD CHANGE THE OUTCOME

Technical capability is only one link in the chain from model performance to employment. The Organisation for Economic Co-operation and Development (OECD) defines exposure as overlap between a job’s tasks and what AI can theoretically perform; it emphasizes that capability is not the probability of automation. It also identifies displacement, productivity-driven labor demand, and new tasks or jobs as offsetting channels. Adoption is constrained by data access, reliability, liability, cybersecurity, regulation, workflow integration, cost, skills, and acceptance by workers and customers. The ILO describes its potential-exposure figures as an upper-bound scenario conditional on full implementation.

Use is also uneven. In an August 2024 nationally representative U.S. survey, 28% of employed respondents reported using generative AI for work; among users, 75% used it for no more than an hour on a typical work-use day. That evidence does not predict 2027 outcomes, but it shows why a high exposure score should not be treated as proof of immediate replacement. Employers may respond through attrition, redeployment, or output expansion rather than layoffs.

The broad global numbers require particular restraint. WEF’s projection of 170 million jobs created and 92 million displaced between 2025 and 2030 describes structural churn driven by multiple forces—technology, economic conditions, demographics, geopolitics, and the green transition—not AI-only job losses. It is based on surveyed employer expectations and a selected 1.2-billion-worker dataset. It should not be converted into an exact forecast of jobs lost in 2027.

GEOGRAPHIC & DEMOGRAPHIC CAVEATS

Exposure and realized harm will differ sharply by country, industry, and labor-market institution. Across ILO’s four exposure gradients, 34% of employment in high-income countries is exposed, compared with 11% in low-income countries. That gap reflects differences in occupational structure and digitalization; it does not prove that high-income countries will suffer proportionately greater job loss. Infrastructure, skills, wage levels, regulation, firm capability, and the availability of complementary work determine whether exposure becomes substitution.

Risk is also gendered because women are disproportionately represented in exposed clerical work. Globally, 4.7% of female employment is in ILO’s highest exposure gradient, compared with 2.4% of male employment. In high-income countries, the corresponding figures are 9.6% and 3.5%, while 41% of female employment and 28% of male employment fall across all four exposure gradients. These measures describe occupational exposure, not inevitable adverse outcomes for individual women or men. They do, however, identify where hiring slowdowns and transitions could amplify existing labor-market inequalities.

Age is another relevant fault line. The U.S. evidence to date is more suggestive for workers aged 22–25 than for workers over 25. A reduced first-job hiring rate is not equivalent to unemployment: prospective entrants may remain in education, take other work, or leave the labor force. Nevertheless, fewer entry opportunities can matter over time because early jobs build experience, networks, and career progression.

CONCLUSION

By 2027, the occupations most plausibly negatively impacted by AI will be those with a dense block of standardized, digital, and readily auditable tasks: clerical data work, routine financial operations, and parts of customer service and transaction processing. The next tier is likely to include entry-level professional work where AI substitutes for predictable drafts, analysis, code, or document handling. Writers, translators, designers, and broader professional occupations are more likely to face changing assignments, rates, and entry paths than wholesale erasure.

The disciplined conclusion is therefore neither complacency nor inevitability. AI exposure signals a need to monitor and redesign work; it does not by itself forecast unemployment. For 2027, the most credible warning sign is fewer entry-level opportunities and more task consolidation in high-substitution work, especially where demand does not expand enough to absorb productivity gains.