The Displacement Shadow: How Fear of AI Replacement Erodes Workplace Conduct, and How Leaders Can Rebuild Security and Commitment
Download the article here:
Listen to a review of this article:
Abstract: Fear of being replaced by artificial intelligence has become a mainstream workplace experience, and new evidence suggests that it carries consequences beyond anxiety. This article synthesizes research on job insecurity, occupation insecurity, and moral disengagement in the context of AI adoption. A 2026 multi-industry study found that employees who feared their occupation could become obsolete reported more moral disengagement and, in turn, more counterproductive work behavior, with stronger effects in highly AI-exposed industries. Survey data show that worry about AI is widespread, persists among heavy AI users, and is higher in organizations that have adopted AI. The article examines consequences for conduct, safety, creativity, retention, and wellbeing, then presents five evidence-based responses: transparent AI workforce roadmaps, employment security and redeployment pathways, fair process when roles change, worker voice in AI deployment, and an ethical climate that combines strong leader relationships with clear standards. Documented organizational experiences illustrate each response. The article argues that how leaders manage the anticipation of AI-driven change is itself a determinant of performance.
For most of the past three years, public debate about artificial intelligence and work has focused on a single question: how many jobs will AI eliminate? Forecasts have ranged from alarming to reassuring, and the evidence on actual displacement remains mixed. Yet while economists debate the numbers, a different consequence of AI is already visible inside organizations. Employees who believe their work may soon be done by machines are changing how they behave, and not always for the better.
A study published in Occupational Health Science in September 2026 put this consequence in sharp relief. Surveying workers across 18 industries, Probst and colleagues found that employees who feared that their entire occupation could become obsolete were more likely to engage in moral disengagement, the cognitive process through which people convince themselves that behavior they would normally consider wrong is acceptable. Moral disengagement, in turn, was associated with more self-reported counterproductive work behavior, such as arriving late or browsing the web on work time. The pattern was stronger in industries with high exposure to AI, such as finance and insurance, and weaker in industries with low exposure, such as forestry and construction (Probst et al., 2026; Washington State University, 2026). The study does not show that AI adoption causes misconduct, and its outcome measures rely on employees' own reports. It does, however, identify a plausible pathway through which the anticipation of AI-driven change, rather than the change itself, can erode the conduct and commitment on which organizations depend.
The timing matters because the anticipation is widespread and growing. In a 2024 survey of more than 5,000 U.S. workers, 52% said they were worried about how AI might be used in the workplace in the future, and about a third expected AI to lead to fewer job opportunities for them (Pew Research Center, 2025). In early 2026, 18% of U.S. employees told Gallup that it was likely their job would be eliminated within five years because of new technology, automation, or AI, a figure that rose to 23% among those working in organizations that had already adopted AI (Associated Press, 2026). Employers have also become more willing to cite AI as a reason for job cuts. Challenger, Gray & Christmas, which tracks announced U.S. layoffs, attributed more than 112,000 announced job cuts in the first seven months of 2026 to AI, more than in any previous full year it has tracked (Challenger, Gray & Christmas, 2026). Meanwhile, executives at prominent companies have publicly predicted that AI will shrink parts of their workforces. Employees are reading these signals.
The organizational stakes extend well beyond individual anxiety. Decades of research on job insecurity show that perceived threats to employment are associated with lower job satisfaction, weaker commitment, poorer health, and higher turnover intentions (Sverke et al., 2002; Jiang & Lavaysse, 2018). Research on layoffs shows that the way organizations manage change shapes how both departing and remaining employees respond (Brockner et al., 1994; Datta et al., 2010). The new evidence on AI-related occupation insecurity adds a further, less discussed risk: that fear can weaken the moral self-regulation that keeps everyday conduct on track.
This article examines the emerging evidence on AI-driven insecurity and its consequences, and offers leaders a practical framework for responding. It begins by defining the key concepts and describing the prevalence and distribution of AI-related insecurity. It then examines consequences for organizations and individuals. The core of the article describes five evidence-based responses, illustrated by the documented experiences of organizations across technology, finance, retail, telecommunications, professional services, and entertainment. It closes with recommendations for building long-term workforce security in an era of continuous technological change. The central argument is simple. Leaders cannot eliminate uncertainty about AI, but they can shape how employees experience it, and that experience has measurable effects on behavior.
The AI Occupation Insecurity Landscape
Defining Job Insecurity, Occupation Insecurity, and Moral Disengagement
Job insecurity refers to an employee's perceived threat to the continuity and stability of their current employment (Shoss, 2017). Researchers distinguish between cognitive job insecurity, the perceived likelihood of losing one's job, and affective job insecurity, the worry or anxiety that accompanies that perception; meta-analytic evidence suggests that the two forms have related but distinct consequences (Jiang & Lavaysse, 2018). Job insecurity is a subjective experience. Two employees in the same organization, facing the same objective risk, may experience very different levels of insecurity depending on their personal circumstances, information, and relationships.
Occupation insecurity extends this concept from a particular job to an entire line of work. Roll et al. (2023) defined it as the fear that one's occupation may disappear or change substantially in content because of automation, and they developed and validated a scale to measure it among more than 1,300 employees in the United Kingdom. Occupation insecurity is distinct from job insecurity because changing employers offers no protection. An accountant who worries that AI will automate much of accounting cannot solve the problem by moving to a different accounting firm. This distinction matters for leaders, because the reassurance that "your job is safe here" does not address the deeper concern that the skills an employee has built over a career are losing their value.
Related research has examined employees' awareness of smart technology, AI, robotics, and algorithms, sometimes abbreviated as STARA awareness. In an early study of employees in New Zealand, those who were more aware of the potential for these technologies to replace their jobs reported lower organizational commitment and career satisfaction and higher turnover intentions (Brougham & Haar, 2018). The finding anticipated the broader pattern documented in more recent work: the perception of technological threat can affect attitudes and behaviors well before any actual displacement occurs.
Moral disengagement is the set of cognitive mechanisms that allow people to behave in ways that conflict with their own moral standards without experiencing the self-condemnation that would normally restrain them (Bandura, 1999). Bandura identified mechanisms such as moral justification, euphemistic labeling, advantageous comparison, displacement or diffusion of responsibility, distortion of consequences, and blaming or dehumanizing those affected. In organizational settings, moral disengagement predicts unethical decision making (Detert et al., 2008) and a range of unethical behaviors rated by supervisors and coworkers (Moore et al., 2012). An employee who thinks, "The company is going to replace me with AI anyway, so why should I go the extra mile?" or "Everyone is cutting corners now," is engaging in a form of moral disengagement.
Counterproductive work behavior refers to voluntary behavior that harms, or is intended to harm, the organization or its members. It includes withdrawal behaviors such as lateness and time wasting, production deviance, misuse of resources, and interpersonal behaviors such as incivility toward coworkers. Counterproductive behavior is often subtle and cumulative. No single instance of extended web browsing will damage an organization, but a widespread shift in norms can.
Finally, AI exposure refers to the degree to which the tasks in an occupation or industry overlap with the capabilities of AI. Felten et al. (2021) developed widely used measures of occupational, industry, and geographic exposure by linking advances in AI applications to the abilities required by different occupations. Eloundou et al. (2024) estimated that about 80% of the U.S. workforce could have at least 10% of their tasks affected by large language models, and about 19% could see at least half of their tasks affected. Exposure is not the same as replacement. Many exposed tasks may be augmented rather than automated, and adoption depends on cost, regulation, and organizational choices. Exposure does, however, shape how plausible the threat of replacement feels to employees.
Prevalence, Drivers, and Distribution
AI-related worry is now a mainstream experience among U.S. workers. In a 2023 American Psychological Association survey of more than 2,500 employed adults, 38% said they worried that AI might make some or all of their job duties obsolete (American Psychological Association [APA], 2023). In Pew Research Center's 2024 survey, 52% of workers said they were worried about how AI may be used in the workplace in the future, compared with 36% who felt hopeful, and 32% said AI would lead to fewer job opportunities for them in the long run (Pew Research Center, 2025). Gallup's 2026 data show that while the share of employees who think their own job is likely to be eliminated within five years remains a minority, at 18%, it has risen from 15% a year earlier and is higher, at 23%, in organizations that have adopted AI (Associated Press, 2026).
Worry does not appear to diminish with use. In Workday's August 2026 survey of nearly 6,000 workers, 62% of employees who used AI for nearly all of their work said they believed AI would make their current skills less valuable, even though 76% of the same heavy users expected AI to open new career opportunities (Workday, 2026). This combination of perceived threat and perceived opportunity is important. Employees are not simply resistant to AI. Many are using it intensively while simultaneously wondering what it will mean for the value of what they know.
Several drivers help explain the spread of AI-related insecurity. The first is leadership communication. Over the past three years, chief executives at several large companies have publicly linked AI to reduced hiring or smaller workforces. In 2023, IBM's chief executive said that roughly 30% of about 26,000 non-customer-facing roles could be replaced by AI and automation over five years (Data Center Knowledge, 2023). In 2025, Amazon's chief executive told employees that he expected generative AI to reduce the company's corporate workforce over the next few years, and later that year the company announced about 14,000 corporate job cuts (FOX 5 DC, 2025). Statements like these are often intended for investors, but employees hear them too.
The second driver is the visibility of AI-attributed job cuts. Challenger, Gray & Christmas data show that AI became one of the most frequently cited reasons for announced job cuts in 2026, with more than 87,000 AI-attributed cuts in the first five months of the year and more than 112,000 through July (Challenger, Gray & Christmas, 2026; HRD, 2026a). These figures reflect employers' stated reasons for announced cuts rather than a measured causal effect of AI on employment, but they shape the narrative employees encounter in the news.
The third driver is uncertainty about career paths. Workday's 2026 analysis found that internal moves declined at 57% of employers year over year, nearly four in ten employees reported experiencing a reorganization or restructuring in the past year, and while 79% of workers said they knew which skills they needed, only 66% said their employer was helping them develop those skills (Workday, 2026). When employees see fewer visible paths forward and limited support for building new skills, the threat of AI feels less like a challenge to be met and more like a fate to be endured.
AI-related insecurity is not evenly distributed. The Probst et al. (2026) study found that the pathway from occupation insecurity to moral disengagement and counterproductive behavior was stronger in industries with high AI exposure, such as finance and insurance, than in industries with low exposure (Washington State University, 2026). Research on job insecurity more generally shows that its effects vary with economic vulnerability, such as dependence on a single income, and with psychological vulnerability (Shoss, 2017), and meta-analytic evidence suggests that some groups suffer more from insecurity than others (Cheng & Chan, 2008). Workers in roles centered on routine information processing, such as customer service, back-office administration, and some forms of content production, have been the focus of many of the most visible AI-related announcements, which likely heightens insecurity in those roles.
Organizational and Individual Consequences
Organizational Performance Impacts
The most immediate organizational consequence of AI-related insecurity is erosion of discretionary effort and everyday conduct. The Probst et al. (2026) findings suggest a mechanism. When employees fear that their occupation is disappearing, some appear to loosen their moral self-regulation, justifying behaviors such as lateness, time wasting, or misuse of resources that they would otherwise consider inappropriate. The authors suggest that employees may rationalize cutting corners or undermining others as a form of self-protection or as compensation for a perceived injustice (Washington State University, 2026). The study measured these behaviors through employees' own reports, which may understate their true frequency, since people tend to underreport behavior they view as undesirable.
These findings are consistent with earlier research on job insecurity. In two time-lagged studies of Chinese employees, Huang et al. (2017) found that job insecurity increased both organizational deviance, such as wasting resources, and interpersonal deviance, such as rudeness toward colleagues, as well as intentions to quit, and that these effects operated through moral disengagement. Probst and colleagues have also documented that job insecurity is associated with lower safety motivation and compliance and with more accidents and injuries (Probst & Brubaker, 2001), effects that depend on the strength of the organization's safety climate (Probst, 2004). A two-wave study of U.S. workers found that job insecurity predicted later safety-related moral disengagement, which in turn predicted poorer safety performance (Probst et al., 2020). For organizations in safety-sensitive industries, AI-related insecurity is therefore not only a morale issue but also a potential safety risk.
Job insecurity can also reshape where employees direct their effort. Research by Probst et al. (2007) found that job insecurity was associated with gains in some forms of productivity but losses in creativity, suggesting that insecure employees may focus on visible, measurable output at the expense of experimentation and innovation. This is a particular concern in the context of AI adoption, which depends heavily on employees' willingness to experiment with new tools and redesign their own work. Employees who fear that demonstrating AI's usefulness will hasten their own replacement have little incentive to share discoveries about how AI can do their jobs.
Insecurity also affects retention, often in ways that harm organizations most. Meta-analytic research shows that job insecurity is associated with higher turnover intentions (Sverke et al., 2002), and in APA's 2023 survey, 46% of workers who worried about AI making their duties obsolete said they planned to look for a new job within the next year, compared with 25% of those who were not worried (APA, 2023). The employees most able to leave are often those with the strongest skills and best outside options, and Huang et al. (2017) found that the link between insecurity and deviance through moral disengagement was strongest when employees had attractive alternatives elsewhere. Organizations can therefore lose both the conduct of those who stay and the talent of those who go.
Finally, poorly managed AI transitions can damage the psychological contract, the set of beliefs employees hold about what they owe their employer and what their employer owes them. Meta-analytic evidence shows that perceived breach of the psychological contract is strongly associated with mistrust and feelings of violation and moderately associated with lower commitment and higher turnover intentions (Zhao et al., 2007). When employees have been encouraged to adopt AI and then see AI cited as a reason for layoffs, many will interpret the sequence as a breach.
Individual Wellbeing and Stakeholder Impacts
For employees, AI-related insecurity is associated with a range of wellbeing harms. In APA's 2023 survey, workers who were worried about AI making their duties obsolete were more likely than those who were not worried to say that work had a negative impact on their mental health (51% versus 29%), to report feeling tense or stressed during the workday (64% versus 38%), and to say they felt they did not matter to their employer (41% versus 23%) (APA, 2023). They were also more likely to report irritability or anger toward coworkers or customers and to say they lacked motivation to do their best. These are correlational findings, and worry about AI may coexist with other workplace stressors, but they align with broader meta-analytic evidence linking job insecurity to poorer mental and physical health (Sverke et al., 2002; Cheng & Chan, 2008).
The wellbeing effects of occupation insecurity may be especially difficult because they touch on identity. A professional who has spent years developing expertise may experience the prospect of that expertise becoming obsolete as a threat to self-worth, not just to income. Roll et al. (2023) found that occupation insecurity was associated with burnout and lower work engagement. Unlike job insecurity, which can sometimes be addressed by finding a new employer, occupation insecurity may require a person to reimagine their career, a much more daunting prospect, particularly for mid- and late-career workers.
The consequences also extend to customers and other stakeholders. Counterproductive behavior, reduced helping, and irritability toward others can degrade service quality. Klarna, the Swedish payments company, offers a cautionary example of how AI-driven workforce decisions can affect customers. In early 2024 the company reported that its AI assistant was handling two-thirds of customer service chats and doing work equivalent to that of 700 full-time agents (Fortune, 2024). By May 2025, however, its chief executive acknowledged that cost had been too dominant a factor in its approach and that quality had suffered, and the company began recruiting human agents so that customers could always reach a person (Fortune, 2025a). The Klarna case was primarily about service quality rather than employee conduct, but it illustrates that AI transitions designed around cost alone can have consequences that leaders did not anticipate.
Evidence-Based Organizational Responses
The evidence reviewed above suggests that AI-related insecurity harms organizations through several pathways: reduced commitment and effort, moral disengagement and counterproductive behavior, safety risks, loss of talent, and damage to the psychological contract. The good news is that the same body of research identifies factors that can weaken these pathways. Insecurity is shaped by what leaders say and do, by the fairness of how changes are made, by the opportunities employees see for themselves, by whether they have a voice, and by the ethical climate and relationships around them. Table 1 summarizes five evidence-based responses.
Table 1: Evidence-based organizational responses to AI-related occupation insecurity
Response | Evidence Base | Key Practices | Illustrative Examples |
Transparent, credible AI workforce roadmaps | Job insecurity research; psychological contract breach (Shoss, 2017; Zhao et al., 2007) | Explain intended uses of AI and their workforce implications; avoid investor-only messaging; update regularly; acknowledge uncertainty honestly | Shopify; Duolingo; Walmart; IBM |
Employment security commitments and redeployment pathways | Occupation insecurity and job insecurity research (Roll et al., 2023; Sverke et al., 2002) | Redeployment-first policies; reskilling linked to real roles; internal mobility windows; use of attrition before layoffs | IKEA; AT&T; Amazon; BT Group; JPMorgan Chase |
Fair process when roles change | Procedural justice and layoff research (Brockner et al., 1994; Datta et al., 2010; Greenberg, 1990) | Clear criteria; advance notice; adequate explanation; respectful treatment; generous transition support | Dropbox; Accenture |
Worker voice and co-design of AI deployment | Psychological contract and job insecurity research (Shoss, 2017; Zhao et al., 2007) | Involve employees and representatives in deployment decisions; agree guardrails; share information early | Microsoft and the AFL-CIO; Writers Guild of America; Klarna |
Ethical climate, leader relationships, and accountability | Moral disengagement and job insecurity research (Huang et al., 2017; Moore et al., 2012; Probst et al., 2020) | Strengthen manager relationships; reinforce clear ethical and safety standards; address moral disengagement directly | Zerodha; Salesforce |
Transparent, Credible AI Workforce Roadmaps
Uncertainty is a central ingredient of insecurity. Job insecurity is, by definition, a perception about future threats (Shoss, 2017), and in the absence of credible information, employees tend to fill gaps with the most alarming interpretations available, often drawn from media coverage or from executives' statements to investors. Research on psychological contract breach suggests that what employees experience as broken promises is especially damaging (Zhao et al., 2007), which means that inconsistency between what leaders say internally and what they do, or say externally, can be more harmful than difficult news delivered honestly. Transparency does not require leaders to promise what they cannot deliver. It requires them to explain how they intend to use AI, what that is likely to mean for different kinds of work, what they do not yet know, and how decisions will be made.
Effective approaches:
Role-level clarity: Describe which tasks and workflows the organization intends to automate or augment, and which roles are likely to grow, shrink, or change, rather than speaking only in general terms.
One story for all audiences: Ensure that messages to employees, investors, and the media are consistent, because employees read all three.
Honest uncertainty: Acknowledge what leaders do not yet know and commit to a timetable for sharing more.
Explicit decision rules: Explain the principles that will guide workforce decisions, such as whether redeployment will be attempted before layoffs.
Regular updates: Revisit the roadmap at predictable intervals so that employees are not dependent on rumor.
Shopify, the e-commerce technology company, illustrates the power of explicit rules. In April 2025 its chief executive publicly shared an internal memo stating that, before asking for more headcount or resources, teams must demonstrate why they cannot get the work done using AI (TechCrunch, 2025). The memo was clear about expectations, which reduces one kind of uncertainty, but it also signaled that AI would shape hiring, which may have heightened concern among some employees about future opportunities.
Duolingo, the language learning company, shows how messages can be misread when context is missing. In April 2025 its chief executive posted an "AI-first" memo stating that the company would gradually stop using contractors for work AI could handle. The memo drew public backlash and fears of layoffs. Months later, he said that he had not given enough context and that the company had never laid off full-time employees and did not plan to (Fortune, 2025b). The episode shows that how leaders frame AI strategy can shape employees' and the public's sense of threat as much as the strategy itself.
Walmart offers a contrasting framing. In 2025 its chief executive said that AI would change essentially every job at the company, while a spokesperson said that the company expected global headcount to remain roughly flat over the following three years even as the mix of roles changed. The company also created new AI-related roles and partnered with a certification program intended to help workers build AI skills (Fortune, 2025c). By pairing an honest acknowledgment of change with a specific statement about headcount and visible investments in skills, Walmart gave employees more information to work with than a general statement about AI's potential would have provided.
IBM's 2023 experience illustrates a common pitfall. After its chief executive told an interviewer that hiring for some back-office roles could slow or pause and that a substantial share of non-customer-facing roles could be replaced by AI over five years, a company spokesperson later clarified that there was no blanket hiring pause (Data Center Knowledge, 2023). Statements made for external audiences can create internal uncertainty that requires clarification. The pitfall for leaders is to assume that strategic statements about AI's potential are understood by employees as long-term possibilities rather than near-term threats. Communication about AI's workforce implications should be planned with employees as a primary audience.
Employment Security Commitments and Redeployment Pathways
Transparency reduces uncertainty, but employees also need reasons to believe that the organization will help them through change. Research on occupation insecurity suggests that the core fear is not losing a particular job but losing the value of one's skills and career (Roll et al., 2023). The most direct response is to create credible pathways from roles that are shrinking to roles that are growing, backed by investment in reskilling and, where feasible, by commitments to attempt redeployment before layoffs. The research on job insecurity suggests that its harms stem from perceived threat (Sverke et al., 2002); credible pathways can reduce that perception even when roles are changing substantially.
Effective approaches:
Redeployment first: Commit to attempting internal redeployment before external layoffs when roles are affected by AI, and report on how often redeployment succeeds.
Reskilling tied to real roles: Link training to specific internal openings and career paths rather than offering generic courses with no clear destination.
Attrition-based adjustment: Where workforce reductions are necessary, rely as much as possible on natural attrition and voluntary programs.
Protected learning time: Give employees paid time to build new skills, rather than expecting them to retrain on their own time.
Transparent mobility data: Publish internal mobility rates and success stories so that pathways are visible and credible.
IKEA, the home furnishings retailer, offers one of the most frequently cited examples of redeployment. After introducing an AI assistant to handle routine customer inquiries, the company retrained thousands of call center workers, reported as about 8,500, to work as remote interior design advisers, rather than eliminating their roles. The company has reported that its remote sales channel has since become a significant source of revenue (People Matters, 2026). The case shows how AI that automates routine tasks can create capacity for higher-value work if leaders choose to redeploy rather than reduce.
AT&T's earlier reskilling initiative, launched well before the current wave of generative AI, demonstrates the scale of investment that large transitions can require. Facing a mismatch between its workforce's skills and its future technology needs, the company invested heavily in retraining a substantial share of its employees, through online courses developed with education partners and tuition support for advanced degrees (Pressman, 2017). The initiative showed both the potential and the difficulty of large-scale reskilling, which depends on employees' willingness to invest their own time and on the availability of roles for them to move into.
Amazon's 2025 job cuts illustrate a more limited form of redeployment. When the company announced about 14,000 corporate job cuts, it gave most affected employees 90 days to search for an internal role, alongside severance and outplacement support for those who did not find one (FOX 5 DC, 2025). BT Group, the British telecommunications company, announced in 2023 that it planned to reduce its workforce by up to 55,000 by 2030, with around 10,000 of those reductions expected to come from digitization, automation, and AI, and said it would work with its union partners and rely on natural attrition where possible (Associated Press, 2023). These approaches signal some commitment to employees, but the scale of the reductions also shows that redeployment cannot always prevent job losses.
A common pitfall is to offer reskilling without destinations. When employees are encouraged to take courses but see no corresponding change in hiring practices or internal mobility, reskilling programs can come to feel like a box-checking exercise or even a prelude to layoffs. Workday's 2026 finding that employees' awareness of the skills they need outpaces their employers' support for developing them (Workday, 2026) suggests that many organizations have not yet closed this gap.
JPMorgan Chase, operating in one of the industries that Probst et al. (2026) identified as most exposed to AI, has made redeployment an explicit part of its public AI strategy. In his 2024 shareholder letter, its chief executive said the bank expected AI to reduce some job categories while creating others and pledged to retrain and redeploy affected employees, and in 2026 he described the bank's redeployment plans as large and in need of further expansion. The bank's own figures show the shift under way, with operations and support functions shrinking modestly in 2025 while revenue-producing and front-office support teams grew (HRD, 2026b). For employees in a highly exposed industry, a visible, sustained redeployment commitment from the top can be a meaningful counterweight to occupation insecurity, provided that the commitment is reflected in actual internal moves.
Fair Process When Roles Change
Some roles will be eliminated, and no amount of communication or reskilling will change that. When this happens, the research on organizational justice offers some of the most robust guidance available. In three studies of layoff victims and survivors, Brockner et al. (1994) found that unfavorable outcomes provoked strong negative reactions mainly when procedural justice was low; when employees perceived the process as fair, the negativity of the outcome had much less effect. Greenberg's (1990) classic field study found that employee theft rose after a temporary pay cut, but that it rose much less at a plant where managers explained the reasons for the cut thoroughly and respectfully. These findings are directly relevant to the Probst et al. (2026) suggestion that some counterproductive behavior reflects an attempt to compensate for perceived injustice. If employees see AI-related changes as arbitrary or disrespectful, they have more material from which to construct moral justifications for retaliation. A broad review of downsizing research also documents the costs that poorly managed reductions impose on survivors and on organizational performance (Datta et al., 2010).
Effective approaches:
Clear criteria: Explain how decisions about which roles are affected are made, and apply criteria consistently.
Adequate explanation: Provide a thorough, honest explanation of why changes are happening, including the role of AI and other factors.
Respectful treatment: Deliver news in person where possible, with dignity, and avoid impersonal mass notifications.
Advance notice and support: Give affected employees as much notice as possible, along with severance, extended benefits, and career transition support.
Attention to survivors: Communicate with remaining employees about what the changes mean for them, because survivors watch how departing colleagues are treated.
Dropbox, the cloud storage company, announced in April 2023 that it would cut about 500 employees, roughly 16% of its workforce, with its chief executive linking the decision in part to the company's shift toward the AI era. The company offered severance of at least 16 weeks of pay plus additional weeks based on tenure, along with job placement and career coaching support (CNBC, 2023). Whatever one thinks of the decision itself, the transparency about reasons and the substance of the support offered are consistent with what the procedural justice research recommends.
Accenture, the professional services firm, reported in September 2025 that it had exited more than 11,000 people as part of a restructuring program, with its chief executive explaining that employees whose skills could not be updated through reskilling would leave the company, even as the firm planned to grow headcount in AI-related roles (CX Today, 2025). The case highlights an important tension. A firm can invest heavily in reskilling and still conclude that reskilling is not viable for some employees. How that conclusion is reached, communicated, and supported will shape whether remaining employees see the process as fair or as a warning about their own futures.
A common pitfall is to treat fairness as a matter of severance alone. Generous packages matter, but research suggests that the perceived fairness of the decision process and the respectfulness of treatment can matter as much as the size of the outcome (Brockner et al., 1994; Greenberg, 1990). Organizations that announce AI-related reductions abruptly, explain them vaguely, or contradict earlier assurances risk damaging the commitment of the employees they most want to keep.
Worker Voice and Co-Design of AI Deployment
Insecurity is heightened by a sense of powerlessness. Research on job insecurity identifies perceived control as an important factor in how employees respond to threat, with proactive coping representing one of the paths through which insecure employees can respond constructively (Shoss, 2017). When employees, or their representatives, have a voice in how AI is deployed, they gain a measure of control over the changes that affect them, as well as information that reduces uncertainty. Worker voice also improves the quality of AI deployment, because employees often understand the nuances of their tasks better than those designing the systems. Involving employees in decisions can also help protect the psychological contract by making changes something done with employees rather than to them (Zhao et al., 2007).
Effective approaches:
Early involvement: Involve employees and their representatives in decisions about where and how AI will be used before systems are deployed.
Co-designed guardrails: Agree on principles governing AI use, such as human review of consequential decisions, disclosure of AI-generated work, and limits on use for surveillance.
Pilot feedback loops: Use pilots with structured employee feedback before scaling AI tools, and act visibly on what is learned.
Shared information: Give employee representatives access to information about AI plans and their workforce implications.
Formal agreements where appropriate: In unionized settings, negotiate provisions on AI deployment, training, and job security.
Microsoft and the AFL-CIO, the largest federation of U.S. labor unions, announced a partnership in December 2023 that committed the company to share information about AI trends with labor leaders and workers, to incorporate worker perspectives into the development of AI technology, and to work together on policy and skills for frontline workers (Microsoft, 2023). The agreement represents an attempt to give workers a structured voice in AI development at the industry level rather than only within individual workplaces.
The Writers Guild of America secured provisions on AI in its 2023 contract with film and television studios after a lengthy strike. Under the agreement, AI cannot write or rewrite literary material, AI-generated material cannot be treated as source material, companies cannot require writers to use AI, and companies must disclose when materials given to writers have been generated by AI (Writers Guild of America, 2023). The agreement shows how collective voice can translate broad anxiety about AI into specific, enforceable guardrails.
Klarna's experience, described earlier, also offers a lesson in this area. The company's initial deployment of AI in customer service was designed primarily to reduce costs, and its later decision to recruit human agents reflected a recognition that customers valued being able to reach a person (Fortune, 2025a). Earlier involvement of frontline service staff, who understand where customers need human judgment, might have surfaced those quality concerns sooner.
A common pitfall is to confuse communication with voice. Announcing AI plans at a town hall and inviting questions is not the same as involving employees in decisions. Employees quickly recognize consultation that has no influence on outcomes, and such exercises can increase cynicism rather than reduce insecurity.
Ethical Climate, Leader Relationships, and Accountability
The fifth response addresses the mechanism at the heart of the Probst et al. (2026) findings: moral disengagement. Research suggests that the link between insecurity and misconduct is not inevitable and depends on the relationships and climate surrounding employees. Huang et al. (2017) found that the path from job insecurity to deviance through moral disengagement held only when employees had poor-quality relationships with their supervisors or attractive options elsewhere; high-quality leader-member exchange neutralized the effect. Probst et al. (2016) similarly found that leader-member exchange moderated the health and safety consequences of job insecurity. However, the research also contains an important caution. In a study of safety-related moral disengagement, Probst et al. (2020) found that perceived organizational and supervisor support, contrary to expectations, strengthened rather than weakened the indirect effect of job insecurity on poor safety performance through moral disengagement. One interpretation is that support without clear standards may be read as permission. The practical lesson is that strong relationships must be paired with clear and consistently reinforced expectations about ethical and safe conduct. Research on moral disengagement shows that it is a measurable propensity that predicts unethical behavior (Moore et al., 2012) and that it is influenced by factors such as empathy and moral identity (Detert et al., 2008), which suggests that organizations can address it directly.
Effective approaches:
Relationship-focused managers: Equip managers to hold regular, candid conversations with team members about how AI is affecting their work and careers.
Support paired with standards: Combine empathy and support with clear, consistently enforced expectations about conduct, quality, and safety.
Naming rationalizations: Help managers and employees recognize common moral disengagement patterns, such as the belief that misconduct is justified because the organization does not value its people.
Safety climate reinforcement: In safety-sensitive settings, increase attention to safety climate during periods of AI-related change, when insecurity may weaken compliance.
Leader credibility: Ensure that senior leaders' statements about AI and jobs are consistent with their actions, because inconsistency undermines the relationships on which ethical climate depends.
Zerodha, an online brokerage based in India, illustrates how leaders can frame AI adoption as an ethical commitment. In 2023 its chief executive publicly stated that the company would not fire anyone because new technology made their job redundant, said the company had created an internal AI policy to address employee anxiety, and argued that companies using AI as a reason for layoffs to boost profits would worsen inequality (Mollman, 2023). Whether or not other firms can make the same commitment, the example shows a leader addressing employee anxiety directly and grounding AI strategy in stated values, which can strengthen the relational foundation that the research suggests buffers insecurity.
Salesforce, the enterprise software company, illustrates the cost of inconsistency. In 2025 its chief executive said publicly that he did not foresee mass layoffs of white-collar workers due to AI, and about two months later said that the company had reduced its customer support workforce from about 9,000 to about 5,000 because AI agents meant it needed fewer people (UC Today, 2025). For employees across the organization, such a shift can make future assurances harder to believe, weakening the trust that helps employees remain committed during change.
A common pitfall is to treat counterproductive behavior during AI transitions purely as a disciplinary matter. Discipline is sometimes necessary, but if the underlying insecurity and perceived injustice remain unaddressed, punishing individual behaviors may deepen employees' sense that the organization is against them, providing further material for moral disengagement. The more effective approach combines accountability with attention to the conditions that make misconduct seem justified.
Building Long-Term Workforce Security
AI-related insecurity is unlikely to fade quickly. AI capabilities continue to advance, organizations continue to experiment with new uses, and public narratives about job loss will remain prominent. Leaders therefore need not only responses to immediate anxiety but also durable capabilities that sustain security and commitment through continuous change. Three capabilities are especially important.
Skills Investment That Matches the Rhetoric
The first capability is a credible, sustained investment in employee development. Many organizations now describe AI as a reason to upskill their workforces, but the evidence suggests a gap between rhetoric and reality. Workday's 2026 research found a 13-point gap between the share of employees who knew what skills they needed and the share who said their employer was helping them develop those skills, and it found that mentions of training skills in job requisitions had declined (Workday, 2026). Because occupation insecurity is rooted in fear that one's skills are losing value (Roll et al., 2023), investment in skills addresses the threat at its source.
Effective skills investment has several features. It is linked to specific future roles rather than generic learning catalogs. It provides protected time during working hours. It recognizes and rewards learning, for example by considering newly developed skills in internal hiring decisions. And it is sustained over time rather than launched as a one-time initiative. Organizations should also be honest with employees about where reskilling is likely to lead. A promise that every employee can be reskilled for a growing role may not be credible, and an unmet promise may do more damage than a realistic assessment.
Measuring Insecurity as a Leading Indicator
The second capability is measurement. Most organizations track engagement, turnover, and performance, but few systematically measure job or occupation insecurity, even though insecurity predicts many of the outcomes they care about (Sverke et al., 2002; Jiang & Lavaysse, 2018). Validated measures are available, including the Occupation Insecurity Scale developed by Roll et al. (2023), which captures fear that one's occupation may disappear or change substantially because of automation.
Regular measurement of insecurity, broken down by function, role, and level of AI exposure, would allow organizations to identify where anxiety is concentrated and to target communication, reskilling, and support accordingly. Because the Probst et al. (2026) findings suggest that the effects of insecurity on conduct may be stronger in highly exposed areas, insecurity data could also help organizations anticipate where risks to conduct, quality, and safety may rise. Measurement should be paired with action. Employees who report insecurity in surveys and then see no response are likely to conclude that their concerns do not matter, which may deepen the very problem being measured.
Sharing the Gains and Renewing the Psychological Contract
The third capability is a renewed psychological contract for the AI era. The traditional implicit bargain, in which loyalty and performance were exchanged for stable employment, has been eroding for decades, and AI-driven change may accelerate that erosion. Organizations that want employees to embrace AI, share their knowledge of how it can improve work, and maintain high standards of conduct during change need to offer something in return. Research on psychological contract breach shows that unmet expectations are strongly associated with mistrust and reduced commitment (Zhao et al., 2007), and research on job insecurity shows that perceived threats to the employment relationship carry broad costs (Shoss, 2017).
A renewed contract might include commitments such as attempting redeployment before layoffs, sharing a portion of the productivity gains from AI with employees through compensation, time, or development opportunities, involving employees in decisions about AI deployment, and treating employees fairly and respectfully when roles must be eliminated. Not every organization can make every commitment, but every organization can be explicit about what it will and will not promise. The research reviewed in this article suggests that clarity, fairness, voice, and consistency between words and actions are more important than any single policy.
Conclusion
The debate about whether AI will eliminate jobs will continue for years, but organizations do not need to wait for its resolution to experience AI's effects on their workforces. Fear of AI replacement is already widespread (APA, 2023; Pew Research Center, 2025), it is higher where AI has been adopted (Associated Press, 2026), and new evidence suggests that occupation insecurity can erode conduct through moral disengagement, particularly in highly exposed industries (Probst et al., 2026). These findings build on decades of research showing that job insecurity is associated with poorer attitudes, health, safety, and retention (Probst & Brubaker, 2001; Sverke et al., 2002) and that fair process can blunt the harms of difficult outcomes (Brockner et al., 1994). The central lesson for leaders is that the way organizations manage the anticipation of AI-driven change is itself a determinant of performance. Leaders who communicate transparently, create credible pathways, treat people fairly, give employees a voice, and maintain strong relationships alongside clear standards can reduce the shadow that AI casts over their workforces, and in doing so, preserve the trust and conduct on which successful AI adoption depends.
Practical takeaways for leaders:
Tell one consistent story: Communicate AI plans and their workforce implications honestly and consistently to employees, investors, and the public, acknowledging uncertainty where it exists.
Build real pathways: Link reskilling to specific internal roles, provide protected learning time, and attempt redeployment before layoffs wherever feasible.
Make process fair: When roles must change or end, use clear criteria, explain decisions thoroughly, treat people with respect, and support both departing and remaining employees.
Give employees a voice: Involve employees and their representatives in decisions about AI deployment and agree on guardrails before scaling new tools.
Pair support with standards: Strengthen manager relationships while clearly reinforcing expectations about conduct, quality, and safety, and help people recognize rationalizations for misconduct.
Measure insecurity: Track job and occupation insecurity by role and AI exposure, and act visibly on what the data show.
Renew the psychological contract: Be explicit about what the organization will offer employees in return for embracing AI, including a fair share of its gains.
Research Infographic
References
American Psychological Association. (2023). 2023 Work in America survey: Artificial intelligence, monitoring technology, and psychological well-being.
Associated Press. (2023, May 18). UK telecom company BT plans to shed up to 55,000 jobs, replace some with AI. WSLS.
Associated Press. (2026, April 13). As AI use increases at work, many employees still choose not to use it, Gallup poll finds. WTOP.
Bandura, A. (1999). Moral disengagement in the perpetration of inhumanities. Personality and Social Psychology Review, 3(3), 193–209.
Brockner, J., Konovsky, M., Cooper-Schneider, R., Folger, R., Martin, C., & Bies, R. J. (1994). Interactive effects of procedural justice and outcome negativity on victims and survivors of job loss. Academy of Management Journal, 37(2), 397–409.
Brougham, D., & Haar, J. (2018). Smart technology, artificial intelligence, robotics, and algorithms (STARA): Employees' perceptions of our future workplace. Journal of Management & Organization, 24(2), 239–257.
Challenger, Gray & Christmas. (2026, August 6). Challenger report: July 2026.
Cheng, G. H.-L., & Chan, D. K.-S. (2008). Who suffers more from job insecurity? A meta-analytic review. Applied Psychology, 57(2), 272–303.
CNBC. (2023, April 27). Dropbox to cut 500 employees, or 16% of workforce. CNBC.
CX Today. (2025, September 30). Accenture lays off 11,000 staff as part of AI reskilling strategy. CX Today.
Data Center Knowledge. (2023, May 3). IBM could replace 7,800 jobs with artificial intelligence, CEO says. Data Center Knowledge.
Datta, D. K., Guthrie, J. P., Basuil, D., & Pandey, A. (2010). Causes and effects of employee downsizing: A review and synthesis. Journal of Management, 36(1), 281–348.
Detert, J. R., Treviño, L. K., & Sweitzer, V. L. (2008). Moral disengagement in ethical decision making: A study of antecedents and outcomes. Journal of Applied Psychology, 93(2), 374–391.
Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2024). GPTs are GPTs: Labor market impact potential of LLMs. Science, 384(6702), 1306–1308.
Felten, E., Raj, M., & Seamans, R. (2021). Occupational, industry, and geographic exposure to artificial intelligence: A novel dataset and its potential uses. Strategic Management Journal, 42(12), 2195–2217.
Fortune. (2024, February 28). Klarna's AI assistant does the work of 700 full-time staff. Fortune.
Fortune. (2025a, May 9). Klarna turns back to human customer service agents after AI push. Fortune.
Fortune. (2025b, August 18). Duolingo CEO admits controversial AI memo did not give enough context, insists company never laid off full-time employees. Fortune.
Fortune. (2025c, September 30). Walmart CEO Doug McMillon can't think of a single job that won't be changed by AI. Fortune.
FOX 5 DC. (2025, October 28). Amazon layoffs: What we know. FOX 5 DC.
Greenberg, J. (1990). Employee theft as a reaction to underpayment inequity: The hidden cost of pay cuts. Journal of Applied Psychology, 75(5), 561–568.
HRD. (2026a, June 9). AI emerges as most-cited reason for job cuts. HRD America.
HRD. (2026b, February 24). Can you really reassign staff as AI takes their jobs? HRD America.
Huang, G.-H., Wellman, N., Ashford, S. J., Lee, C., & Wang, L. (2017). Deviance and exit: The organizational costs of job insecurity and moral disengagement. Journal of Applied Psychology, 102(1), 26–42.
Jiang, L., & Lavaysse, L. M. (2018). Cognitive and affective job insecurity: A meta-analysis and a primary study. Journal of Management, 44(6), 2307–2342.
Microsoft. (2023, December 11). AFL-CIO and Microsoft announce new tech-labor partnership on AI and the future of the workforce [Press release].
Mollman, S. (2023, May 12). CEO vows not to "fire anyone because of A.I.," warns that others doing so "isn't a good outcome for humanity." Fortune.
Moore, C., Detert, J. R., Treviño, L. K., Baker, V. L., & Mayer, D. M. (2012). Why employees do bad things: Moral disengagement and unethical organizational behavior. Personnel Psychology, 65(1), 1–48.
People Matters. (2026, August 4). IKEA uses AI to reskill 8,500 employees instead of cutting jobs. People Matters.
Pew Research Center. (2025, February 25). U.S. workers are more worried than hopeful about future AI use in the workplace.
Pressman, A. (2017, March 13). Can AT&T retrain 100,000 people? Fortune.
Probst, T. M. (2004). Safety and insecurity: Exploring the moderating effect of organizational safety climate. Journal of Occupational Health Psychology, 9(1), 3–10.
Probst, T. M., & Brubaker, T. L. (2001). The effects of job insecurity on employee safety outcomes: Cross-sectional and longitudinal explorations. Journal of Occupational Health Psychology, 6(2), 139–159.
Probst, T. M., Jiang, L., & Graso, M. (2016). Leader–member exchange: Moderating the health and safety outcomes of job insecurity. Journal of Safety Research, 56, 47–56.
Probst, T. M., Petitta, L., Barbaranelli, C., & Austin, C. (2020). Safety-related moral disengagement in response to job insecurity: Counterintuitive effects of perceived organizational and supervisor support. Journal of Business Ethics, 162(2), 343–358.
Probst, T. M., Petitta, L., Roll, L., Lindgren, R. J., Barbaranelli, C., & Ghezzi, V. (2026). In the shadow of AI: How industry exposure to AI and occupation insecurity fuel moral disengagement and counterproductive work behaviors. Occupational Health Science, 10(3), Article 27.
Probst, T. M., Stewart, S. M., Gruys, M. L., & Tierney, B. W. (2007). Productivity, counterproductivity and creativity: The ups and downs of job insecurity. Journal of Occupational and Organizational Psychology, 80(3), 479–497.
Roll, L. C., De Witte, H., & Wang, H.-J. (2023). Conceptualization and validation of the Occupation Insecurity Scale (OCIS): Measuring employees' occupation insecurity due to automation. International Journal of Environmental Research and Public Health, 20(3), Article 2589.
Shoss, M. K. (2017). Job insecurity: An integrative review and agenda for future research. Journal of Management, 43(6), 1911–1939.
Sverke, M., Hellgren, J., & Näswall, K. (2002). No security: A meta-analysis and review of job insecurity and its consequences. Journal of Occupational Health Psychology, 7(3), 242–264.
TechCrunch. (2025, April 7). Shopify CEO tells teams to consider using AI before growing headcount. TechCrunch.
UC Today. (2025, September). AI agents slash 4,000 Salesforce jobs months after CEO downplays AI's risks on jobs. UC Today.
Washington State University. (2026, October 14). Workers who fear AI replacement show worse workplace behavior [Press release]. EurekAlert!
Workday. (2026, October 5). Workday Global Workforce Report: AI is rewriting jobs more than it's cutting them [Press release].
Writers Guild of America. (2023). Summary of the 2023 WGA MBA.
Zhao, H., Wayne, S. J., Glibkowski, B. C., & Bravo, J. (2007). The impact of psychological contract breach on work-related outcomes: A meta-analysis. Personnel Psychology, 60(3), 647–680.
Jonathan H. Westover, PhD, Chief Research Officer (Nexus Institute for Work and AI); Co-Founder & Chief Workforce and Learning Officer (Future State University); Founder & CEO (Human Capital Innovations); Professor of Organizational Leadership & Change (UVU). Read Jonathan Westover's executive profile here.
Suggested Citation: Westover, J. H. (2026). The Displacement Shadow: How Fear of AI Replacement Erodes Workplace Conduct, and How Leaders Can Rebuild Security and Commitment. Human Capital Leadership Review, 40(3). doi.org/10.70175/hclreview.2020.40.3.6
