OUR PERSPECTIVES
Purpose
Discussions, media narratives, and policy debates around technological change and the workforce have historically been characterized by contrasting extremes and wildly varying predictions about future developments. The ongoing debate around the role of automation and generative artificial intelligence (GenAI) is an especially clear example of this, with impassioned arguments that range from “AI is a fad” to “no one will have a job.”
In popular media and many policy circles, the notion that AI-fueled automation will lead to massive job displacement seems to have taken hold, with The New York Times reporting in May 2025 that the AI job apocalypse may already be here for recent college graduates, and an October 2025 report published by the Senate Committee on Health, Education, Labor, and Pensions warning that AI could threaten about 100 million U.S. jobs in the next decade.
On the other hand, up until very recently, the dominant belief among U.S. employers and policymakers was that there aren’t enough U.S. workers, a concern that continues to burn brightly in several sectors. Furthermore, recent data from the New York Federal Reserve Bank does not suggest that AI adoption by employers is inextricably linked to job displacement. Finally, results from SHRM’s 2025 data brief Automation, Generative AI, and Job Displacement Risk in U.S. Employment suggest that – although millions of current U.S. jobs face high automation displacement risk – this set still represents a small fraction of overall U.S. wage/salary employment.
Amid this debate, data on what workers actually believe will happen to their jobs has been surprisingly limited, despite the fact that these beliefs — whether justified or not — are likely to drive key outcomes such as upskilling and reskilling efforts. To better understand these beliefs, SHRM Thought Leadership1 asked participants a series of questions about job displacement expectations within the next five years. This data brief reviews several key findings regarding these expectations and discusses their implications for the world of work.
KEY FINDING NO. 1
A Majority of Respondents Believe That the Chance Their Current Job Will Be Displaced by Automation or Generative AI Is 10% or Less
Although the SHRM 2025 Automation/AI Survey collected survey data from 20,262 individual U.S. workers spread across more than 750 occupations, those workers tended to be concentrated in a much smaller number of occupations, reflecting the fact that some occupations are much larger than others in terms of employment. In fact, the 46 occupations included in the analysis of this data brief collectively accounted for about 45.1% of all survey observations (9,146 respondents).
After identifying the set of occupations to study, a natural first step is to study the distribution of workers’ beliefs in these occupations, including expectations about automation displacement in general and GenAI displacement in particular. Reported in Figure 1, these distributions offer some basic insights into the types of displacement beliefs that are common among people in these 46 occupations.
Perhaps the most fundamental finding stemming from these results is that a significant minority of survey respondents in these occupations believe that there is a 0% chance that automation in general (36.9% of respondents) or GenAI in particular (39% of respondents) will displace their current job within the next five years. Furthermore, the next most common choice for both probabilities was a value greater than 0 but less than or equal to 10%, meaning that a clear majority of respondents across these 46 occupations placed the likelihood of their job being displaced by automation or GenAI at 1 in 10 or less.3
Even so, it is also true that a notable number of respondents have significantly elevated fears of job displacement via automation and/or GenAI, including a small fraction of respondents who place their odds of displacement at greater than 50%. In short, although most respondents feel confident that their job is unlikely to be replaced by these technological channels in the next five years, it is also true that a subset of respondents have displacement concerns mirroring the dire predictions that capture news headlines. As will be shown below, the types of respondents who expressed these beliefs are disproportionately concentrated in specific occupations.
Related Article
Tailor AI Adoption Strategies to Meet Workforce Needs
Given the report’s finding that most workers expect low displacement risk but some have heightened fears, this resource explains how organizations can customize AI rollouts. Learn to address pockets of concern, foster upskilling, and ensure technology implementation matches workforce readiness.
Related Event
AI+HI Project 2026
Participate in this event to explore the intersection of artificial intelligence and human intelligence. Gain actionable insights for addressing worker concerns highlighted in the report, and learn how organizations can leverage both technology and human expertise to navigate evolving job displacement risks.
KEY FINDING NO. 2
Seven of the Occupations Studied Have a Median Expected Probability of Displacement via Automation of at Least 20%
Figure 2 lists the top occupations as ranked by median expected probability of job displacement by automation. Though the original intent was to discuss a list of the top 10, due to ties there are 11 occupations listed. In any case, one especially striking feature of this list is the dominating presence of occupations in computer and mathematical occupations. In fact, occupations from this group account for six of the 11 occupations listed, including five of the seven with a median expected probability of job displacement by automation at or above 20%. What makes this finding so remarkable is that intuition suggests these types of jobs would generally have been viewed as highly shielded from automation as recently as a few years ago, which underscores how rapidly the technological landscape is shifting.
Although computer and mathematical occupations clearly dominate Figure 2, a handful of additional fields make an appearance. Among these are financial and investment analysts as well as construction and building inspectors, both of which have a median expected probability of job displacement by automation at or above 20%. Once again, it seems unlikely that either occupation would have rated as having high exposure to automation a decade ago; however, advances in technology increasingly made highly complex but routinized tasks (which feature prominently in both occupations) automatable.
The bottom four occupations on this list stand out from the top seven due to a significant downward shift in median expected probability of job displacement by automation (from at least 20% to between 11% and 12%). In other words, respondents in these final four occupations have above-average beliefs that automation will displace their job but have much more conservative projections than respondents in the top seven occupations. Three of these four occupations are in management or business and financial operations roles, including two occupations in human resources specifically. As was the case for occupations that ranked higher on this list, these four occupations had limited-to-no historical association with automation exposure, further underscoring the extent to which emerging technologies have transformed beliefs about what is automatable.
Related Specialty Credential
AI+HI Specialty Credential
Stay ahead as AI transforms the workplace. Master the intersection of technology and human insight with this credential, gaining skills to lead strategic change, address automation and GenAI risks, and drive business results. Empower your career to thrive—take the next step in shaping the future of work.
Related Article
AI and Automation: Real Tips from Real CEOs
Discover practical strategies and firsthand insights from CEOs on navigating workforce automation, directly addressing concerns raised by workers about job displacement risk.
KEY FINDING NO. 3
Five of the Occupations Studied Have a Median Expected Probability of Displacement via Generative AI of at Least 20%
Figure 3 completes our analysis by listing the top occupations studied as measured by median expected probability of job displacement by GenAI. As in Figure 2, the presence of ties means that more than 10 occupations are listed (in this case, there are 15).
The findings reported here mirror much of what we discussed in connection with Figure 2. For example, occupations in the computer and mathematical group are significantly overrepresented, including four of the five occupations with a median expected probability of at least 20%.
Outside of computer and mathematical occupations, the remainder of the list includes one management occupation (human resources managers), one construction occupation (construction and building inspectors), two office and administrative support occupations (customer service representatives and human resources assistants), and four business and financial operations occupations (financial and investment analysts, business operations specialists – all other, accountants and auditors, and financial specialists – all other).
To reiterate a continuing theme, all of the occupations listed in Figure 3 have limited or no historical association with job displacement via technological change; however, many are now prime candidates because they center heavily on routinized tasks that GenAI tools are increasingly proficient at managing. For example, computer programmers appear to be especially exposed to displacement via GenAI, likely because many tools have become highly proficient at producing code for common operations as well as debugging existing code. These advances don’t mean that computer programmers will become entirely obsolete. Instead, it seems likely that GenAI tools will allow a smaller number of human programmers to be much more productive.
Related Article
The New Age of Workforce Planning
Rapid shifts in worker beliefs, especially among computer and mathematical occupations, call for new planning approaches. This resource details how CHROs are moving from headcount to skills-based strategies, helping organizations proactively prepare for evolving job displacement expectations.
Related Article
AI, Job Disruption, and Preparing for What’s Next: A Conversation with Futurist Bugge Hansen
With the report showing that beliefs about automation and GenAI displacement vary by occupation, this conversation explores future workforce scenarios. Gain actionable advice for HR leaders to support adaptability, address uncertainty, and guide employees through ongoing technological transformation.
CONCLUSION
In an environment characterized by sensational headlines pushing wildly varying views on AI and the future of work, this data brief sought to understand the subjective beliefs of workers about job displacement risk in a subset of 46 occupations for which data from the SHRM 2025 Automation/AI survey provided substantial direct evidence.
Our findings suggest that most workers believe that there is a low or even negligible (i.e., 0% to 10%) chance that their current role will be displaced by either automation or GenAI in the next five years. However, there is also a notable subset of workers who set one or both probabilities much higher. Our examination of the median expected probabilities of displacement via automation and GenAI suggests that workers in the computer and mathematical occupational group are especially likely to view displacement as a plausible outcome, though workers in some business and financial operations as well as office and administrative support occupations express similar expectations.
These findings reinforce that workers’ concerns about job displacement through automation and GenAI are real, especially among white-collar workers whose jobs revolve around routinized, computer-based tasks. Whether these concerns reflect actual displacement risk is an open question, though our own preferred measure of occupation-level automation displacement risk introduced in Automation, Generative AI, and Job Displacement Risk in U.S. Employment is strongly correlated with the workers’ beliefs presented here.
Given the current pace of technological change as well as broad uncertainty surrounding labor market and overall economic conditions, it seems overwhelmingly likely that workers, policymakers, and organizations will exhibit dynamic beliefs about the role of automation and GenAI in the future of work. This volatility underscores the need to track these beliefs carefully and proactively identify optimal policies and business practices that help organizations and workers succeed in a rapidly evolving landscape.
Definitions
A key challenge when discussing this topic is clearly defining terms. For instance, automated processes have become so ubiquitous in day-to-day life that it is sometimes easy to forget that such processes often represent critical work tasks. Similarly, AI technologies can appear in subtle ways that may not immediately be associated with AI, such as improved editing features in word processing software. To avoid confusion, we defined the following terms for participants in the SHRM 2025 Automation/AI Survey:
Automation — the technique of making an apparatus, a process, or a system operate autonomously (i.e., without human intervention). An example is manufacturing processes that complete routinized tasks without human input. A task is considered automated if it is completed using automation technology. The degree to which an occupation is automated depends on the extent to which tasks completed within that occupation are done via automation.
Artificial intelligence (AI) — any technology that can independently perform tasks that typically require human intelligence. One example is self-driving vehicles.
Generative AI (GenAI) — any AI technology that leverages training on an archive of content to generate brand-new, unique content (e.g., text or images). One example is ChatGPT, which generates entirely novel and unique content based on a massive database of existing content that the software has been trained on.
To simplify the language of this data brief and provide greater clarity to readers, the following definitions will also be useful:
Job displacement — the elimination of a human job (i.e., a specific position within an organization). Job displacement can occur for many reasons (e.g., organizational restructuring), but in the context of this research, our focus is job displacement stemming from automation and GenAI. In this kind of displacement, jobs are eliminated because technological change has made them unnecessary within an organization. This could take many forms, including (but not limited to) human jobs being literally replaced by machines or a piece of software (e.g., cashiers being replaced with self-checkout kiosks) and jobs being eliminated because higher productivity has reduced the number of people needed to perform a given activity (e.g., a tech company might eliminate some percentage of its computer programmers because GenAI tools have made generating programming code much more efficient and rapid).
Expected probability of job displacement via automation — this value represents each individual survey respondent’s self-reported probability that their current position will be displaced by automation within the next five years (as measured at the time of the survey). When discussing these probabilities for individual occupations, we generally focus on the unweighted median value observed across all survey respondents in the occupation.
Expected probability of job displacement via GenAI — this value represents each individual survey respondent’s self-reported probability that their current position will be displaced by GenAI within the next five years (as measured at the time of the survey). When discussing these probabilities for individual occupations, we generally focus on the unweighted median value observed across all survey respondents in the occupation.
High automation level — we define a job as having a high automation level if at least half of the tasks in that job are automated. A central value that we estimate in our results is the share of employment in a given occupation (or collection of occupations) that meets this threshold. For example, we estimate that about 39.7% of employment in the software developers occupation is highly automated (i.e., at least half of the tasks are automated).
Nontechnical barrier to automation displacement — a non-technical barrier to automation displacement is any barrier to displacement that does not relate to a limitation of the automation technology. In other words, nontechnical barriers to automation displacement can exist even when the technology exists to completely automate a given job. For example, client preferences for interpersonal interaction might present a barrier to displacing employment through automation.
Methods and Data
A detailed review of the SHRM 2025 Automation/AI Survey’s methodology for estimating the share of employment in individual occupations that meet certain conditions (e.g., highly automated with no nontechnical barriers to displacement) is provided in the methodological appendix for Automation, Generative AI, and Job Displacement Risk in U.S. Employment. This specific data brief does not cover that analysis. Instead, it leverages two questions appearing in the same survey that have not been examined to date:
- Regardless of whether or not you are still working in the position, how likely do you think it is that your employer will eliminate your current job within the next five years because automation has made the role obsolete?
- Regardless of whether or not you are still working in the position, how likely do you think it is that your employer will eliminate your current job within the next five years because generative AI has made the role obsolete?
In both cases, respondents selected the percentage probability (from 0% to 100%) using a slider. For individual occupations, we measured unweighted median probabilities reported by respondents within the occupations. Because we needed adequate sample sizes to estimate these medians accurately, our analyses were limited to the 46 occupations for which we had at least 100 individual survey respondents. Although these occupations represent a small share of the 800-plus occupations in the Standard Occupational Classification (SOC) system, they are typically very large occupations by employment level. In the May 2024 BLS OEWS data, these 46 occupations account for about 49.1 million jobs, or nearly 32% of occupational employment in the data.2
Although these 46 occupations do represent a relatively large share of U.S. wage/salary employment, it is important to note that they do not represent a randomly drawn, representative sample of all occupations. In fact, members of certain occupations were significantly more likely to participate in our survey, and, as a result, certain fields tend to be overrepresented in the 46 occupations for which we had at least 100 survey respondents. Even so, the occupations discussed in this brief span 14 of the 22 major civilian occupational groups in the SOC system and include a mix of white-collar, blue-collar, and service roles. Therefore, although these findings cannot be taken to reflect a nationally representative picture of workers’ beliefs about the probability of automation and GenAI job displacement risk, they do provide an instructive look at the types of occupations that tend to encourage heightened or depressed expectations about these displacement risks.
1. These questions were asked as part of the SHRM 2025 Automation/AI Survey. This survey was also the primary data source used in Automation, Generative AI, and Job Displacement Risk in U.S. Employment, although the focus in that analysis was a different set of questions within the survey.
2. In Automation, Generative AI, and Job Displacement Risk in U.S. Employment, we used a proximity-based method to estimate characteristics such as percentage of tasks automated, percentage of tasks done using GenAI, and presence of nontechnical barriers to job displacement through automation for all 831 detailed (i.e., six-digit SOC code) occupations in the May 2024 BLS OEWS data. For occupations with limited or no direct survey observations, this meant estimating values based on additional data from survey respondents in sufficiently similar occupations. Occupational similarity was determined based on an occupational distance matrix capturing the standard Euclidean distance between any two occupations according to the O*NET generalized work activities module. The intuition behind this approach is that occupations with similar work activities should also have similar exposure to automation and GenAI tools used to complete those activities. Whereas this method is intuitive for estimating the likely characteristics of an occupation, using it to infer what workers’ beliefs are in occupations with insufficient direct survey evidence is conceptually problematic. For this reason, our analysis of workers’ beliefs about displacement risk was limited to the 46 occupations for which we had at least 100 direct survey observations.
3. One important note is that we did not constrain survey respondents to choose a probability for job displacement via GenAI that was less than their chosen probability for displacement via automation; thus, it is possible that a respondent could choose a probability of displacement via GenAI that is higher than their chosen probability of displacement via automation. This choice reflects the attitude that — while GenAI can and often is used to automate tasks — its purposes extend to augmenting tasks and making them more efficient, rather than automating them entirely.