Trust in Freefall: Why Relentless Transformation Is Eroding Confidence in Leaders, and How Organizations Can Rebuild It
Download the article here:
Listen to a review of this article:
Abstract: Employee trust in CEOs, direct managers, and co-workers has fallen to record lows in recent survey data, at a time when three in four employees report living through at least one organizational transformation a year, most often AI implementation and strategy shifts. This article examines the erosion of trust in leaders during sustained transformation, drawing on 2026 survey evidence from Edelman, DDI, and Gallup alongside decades of organizational research on trustworthiness, procedural justice, psychological contracts, employee voice, and trust repair. It documents how low trust undermines strategy execution, discretionary effort, AI adoption, and the upward flow of information, and how its costs fall hardest on the employees closest to implementation. Five evidence-based responses are presented: candid, reasoned communication; procedural justice in restructuring; transparent AI and data governance; equipping managers as trust translators; and voice systems paired with systemic trust repair. Illustrated by organizational examples from technology, financial services, aerospace, education technology, and health care, the article concludes with guidance on measuring trust by level, building trust reserves before disruption, and institutionalizing fairness in change governance.
Ask a room of executives whether their people trust them, and most will say yes, perhaps with a caveat about a difficult quarter. Ask the people three levels down, and the answer is often different. That gap has always existed, but in 2026 it is widening in ways that should worry anyone responsible for leading an organization through change. Edelman's annual Trust at Work study, which surveyed more than 10,000 employees across ten markets, found that trust in "my CEO" fell from 70% to 66% over the past year, trust in co-workers dropped from 82% to 73%, and trust in direct managers slid from 79% to 72%, each a record low in the series (Edelman, 2026). A week later, DDI's Global Leadership Forecast 2027 reported that only 34% of leaders trust their immediate manager and just 31% trust their senior leaders, with the lowest levels found among the managers and individual contributors who actually turn decisions into action (Rhyne et al., 2026).
What makes these numbers urgent is the backdrop. Three out of four employees in the Edelman sample said they had lived through at least one organizational transformation in the previous twelve months, most commonly the implementation of artificial intelligence (AI) and shifts in strategy (Edelman, 2026). Organizations are asking people to accept new tools, new structures, and new expectations at a moment when the reservoir of goodwill that makes such requests bearable appears to be draining. Change requires people to take risks on the word of their leaders. Trust is the willingness to take exactly those risks.
I want to be careful not to overstate the story. Survey trust levels move for many reasons, and the studies discussed here are largely cross-sectional, which means they tell us about associations rather than proving that transformation causes trust to fall. Still, the pattern across independent sources is consistent enough to take seriously. Employees whose experience of transformation was positive report much higher trust in their CEO and manager than those whose experience was negative, and the gap is far larger than any year-on-year swing (Edelman, 2026). That observation carries a hopeful implication: how change is led may matter as much as how much change there is.
This article draws on several decades of organizational research to examine why trust in leaders, managers, and colleagues is eroding during sustained transformation, what that erosion costs, and what actually rebuilds trust once it has been damaged.
The Trust in Leaders Landscape
Defining Trust, Trustworthiness, and Trust Repair
Organizational scholars have converged on a fairly precise definition. Trust is a psychological state in which a person accepts vulnerability based on positive expectations of another party's intentions or behavior (Rousseau et al., 1998). The emphasis on vulnerability is important. Trusting a leader means accepting that the leader could act in ways that harm you, and choosing to proceed anyway because you expect they will not. In a transformation, employees are vulnerable in concrete ways: their roles might be redesigned, their skills devalued, their teams merged, or their work handed to an AI system.
Trust should be distinguished from trustworthiness, the set of qualities in the other party that make trust reasonable. The most influential model, from Mayer et al. (1995), identifies three such qualities. Ability refers to the skills and competence that allow a party to deliver in a particular domain. Benevolence is the extent to which the party is believed to want to do good for the trustor, beyond self-interest. Integrity is the perception that the party adheres to principles the trustor finds acceptable, such as honesty and consistency between words and deeds. A fourth element, trust propensity, captures the trustor's general disposition to trust others. A meta-analysis of 132 independent samples confirmed that ability, benevolence, and integrity each predict trust, and that they also explain variance in risk taking and job performance beyond trust itself (Colquitt et al., 2007). Put simply, people respond to how competent, caring, and principled their leaders appear, not merely to a global feeling of trust.
Two further distinctions matter for this discussion. The first concerns the referent of trust. Employees can trust (or distrust) their direct manager, the senior leadership team, the CEO, their co-workers, and the organization as an abstract entity, and these are different relationships. Dirks and Ferrin's (2002) meta-analysis showed that results differ systematically depending on whether the referent is a direct leader or organizational leadership, with direct leaders emerging as an especially important focus for day-to-day attitudes and behavior. The second distinction is between cognitive trust, grounded in evidence of reliability and competence, and affective trust, grounded in emotional bonds and perceived care. A later meta-analysis of 185 studies found that affective trust was the stronger predictor of independently rated performance and citizenship behavior (Legood et al., 2021).
Finally, a trust violation occurs when one party's actions disconfirm the positive expectations on which trust rested, and trust repair refers to activities that make the trustor's beliefs and intentions more positive after a violation (Dirks et al., 2009; Kim et al., 2004). Closely related is the idea of a psychological contract, the employee's beliefs about the reciprocal obligations between themselves and their employer. When employees perceive that the organization has failed to meet those obligations, they experience psychological contract breach, which Robinson (1996) showed both depends on and erodes trust. Transformation is, among other things, a large-scale renegotiation of psychological contracts, often conducted without anyone saying so out loud.
Prevalence, Drivers, and Distribution
The headline statistics are worth unpacking because the distribution of trust tells a more interesting story than the averages. In the Edelman (2026) data, the first statistically significant year-on-year drop in trust in "my employer" since August 2024 was concentrated in particular markets, with China down 11 points and Germany down 5. Exposure to transformation was nearly universal at the top of organizations: more than 90% of executives, 86% of mid-level managers, and 61% of associates reported experiencing at least one transformation in the prior year. AI implementation was the most frequently cited change (53%), followed by strategy shifts (47%). In several markets, including India and the United Arab Emirates, most employees reported three or more firm-wide changes in a single year.
Figure 1 shows the year-on-year movement in the three interpersonal trust measures. The steepest decline was in trust among co-workers, a nine-point fall, which suggests the erosion is not simply a matter of employees growing skeptical of distant executives. Transformation reshuffles teams, changes who depends on whom, and introduces new competition for a shrinking set of secure roles. All of that can strain lateral relationships as much as vertical ones.
[Insert image: Trust_in_Freefall_Figure_1.png] The experience of transformation is also unevenly distributed. One in three employees told Edelman that their overall transformation experience had not been positive, and the share rose to 47% in Germany and 44% in Canada and Japan. By level, 49% of associates described a less-than-positive experience compared with 26% of mid-level employees and 17% of executives (Edelman, 2026). Associates were roughly three times as likely as executives to have had a poor experience. The people who design change and the people who absorb it are, quite literally, living through different events.
DDI's findings point in the same direction from a different angle. Among executives, 55% trust their own manager and 43% trust senior leaders, but those figures fall steadily down the hierarchy to 25% and 21% among individual contributors (Rhyne et al., 2026). The DDI authors observe that trust is weakest exactly where decisions are converted into action, and that executives may be unaware of the problem because trust can look healthy from the top. Figure 2 displays this gradient.
[Insert image: Trust_in_Freefall_Figure_2.png] Several drivers seem to be at work. The first is sheer economic anxiety. Three in four workers in the Edelman sample feared losing their jobs to some combination of tariffs, automation, offshoring, or inadequate training, and the study identified a gap of more than 21 points between what employees expect on compensation and internal mobility and what employers actually provide (Edelman, 2026). When people feel exposed, they scrutinize leaders' intentions more closely, and benevolence becomes the dimension of trustworthiness under the most pressure.
The second driver is the character of AI-driven change, which is open-ended, technically opaque, and often framed in terms of efficiency that employees reasonably translate as headcount. A global study led by the University of Melbourne with KPMG, surveying more than 48,000 people in 47 countries, found that 58% of employees intentionally use AI at work, yet only 47% had received any AI training and only 40% said their workplace had a policy or guidance on generative AI (University of Melbourne, 2025). Just 46% of respondents globally were willing to trust AI systems. Research on human trust in AI suggests that transparency is especially important for building cognitive trust in embedded and virtual AI, and that early high trust in such systems can decline after use when performance disappoints (Glikson & Woolley, 2020). When organizations deploy tools that employees neither understand nor trust, some of that distrust inevitably attaches to the leaders who chose them.
The third driver is data. Gallup's quarterly workforce study of 22,573 U.S. employees found that only 20% trust their organization's leaders "a great deal" to use employee data responsibly, while 37% have not too much or no trust at all (Kemp & Pendell, 2026). Only 19% of employees said anyone at work had talked with them about the data collected on their work. Six in ten employees in the Edelman (2026) sample feared being monitored without their consent. Data practices have become one more arena in which leaders' integrity is judged, often without any explanation offered.
A fourth driver is the cumulative weight of prior change. Oreg et al.'s (2011) review of six decades of research on change recipients found that reactions depend on pre-change conditions, including the organization's internal context, and on how the change is implemented, not only on its content. Organizations caught in back-to-back transformations can therefore slide into a spiral in which each initiative meets more skepticism than the last.
Organizational and Individual Consequences
Organizational Performance Impacts
The evidence linking trust to performance is unusually robust for organizational research. Dirks and Ferrin's (2002) meta-analysis drew on 106 effect estimates to show that trust in leadership is associated with a range of attitudes and behaviors that matter for organizations, with the strength of relationships depending on whether the referent was a direct or organizational leader. Colquitt et al. (2007) found that trust and the three dimensions of trustworthiness predicted task performance, citizenship behavior, and counterproductive behavior, and that trust was associated with a greater willingness to take risks on behalf of the relationship. Legood et al. (2021) showed that trust in the leader is one of the mechanisms through which leadership styles, particularly servant and ethical leadership, translate into independently rated performance.
The most immediate performance cost of low trust in a transformation is execution failure. DDI's data show that leaders who trust their senior leaders are 5.9 times more likely to say leadership decisions translate into action (Rhyne et al., 2026). The DDI authors describe what happens when trust is absent: managers hedge, or pass decisions along without conviction. Anyone who has watched a strategic initiative wither through a dozen half-hearted cascade meetings will recognize the pattern. DDI also reports that organizations that prioritize trust-building in their leadership development programs meet or exceed financial targets at a 49% higher rate, an association rather than proof of cause, but a striking one.
Edelman's comparison of employees with positive and negative transformation experiences gives a sense of the stakes (Figure 3). Among those with a positive experience, 83% said they would go above and beyond to help the company, compared with 57% of those with a negative experience. The figures for commitment to a long-term career with the employer were 83% and 51%. Perhaps most relevant to AI strategy, 40% of employees with a positive experience said they would embrace increased AI use, versus only 18% of those with a negative experience (Edelman, 2026). In other words, the very transformation that is eroding trust depends on trust to succeed. Organizations that push AI adoption without attending to the experience of change may be undermining their own return on investment.
[Insert image: Trust_in_Freefall_Figure_3.png] Engagement is another channel. ADP has reported that workers who trust their organization's senior leadership are 41 times more likely to be highly motivated and committed (ADP, n.d.). In an October 2026 interview, ADP's chief executive repeated that figure and added that trust in co-workers was associated with a 16-fold increase in strong motivation and commitment (Gerut, 2026). These multipliers come from proprietary research whose methods are not fully described in the public sources, so they are best read as directional. Gallup's data on workplace data trust show a similar alignment: 40% of engaged employees trust leaders a great deal to use data responsibly, compared with 4% of actively disengaged employees (Kemp & Pendell, 2026).
Low trust also degrades the information flow that organizations need to detect problems. When employees doubt that leaders will act on concerns, or fear that they will be punished for raising them, they stay silent. Morrison (2014) describes the decision to speak up as resting on two judgments: whether speaking will make a difference and whether it is safe. Both depend on trust. In a transformation, where the early warning signs of failure are usually visible first to front-line employees, silence can be very expensive. The independent panel that reviewed Boeing's safety culture in 2024, discussed later, offers a sobering illustration of this dynamic in a high-stakes setting (Federal Aviation Administration, 2024).
Finally, trust affects retention, although the relationship is more complicated than often assumed. Zhao et al.'s (2007) meta-analysis of psychological contract breach found strong associations between breach and both feelings of violation and mistrust of management, a strong negative association with job satisfaction, and moderate associations with lower organizational commitment and higher intention to leave, along with modest declines in in-role performance and citizenship. Breach was not associated with actual turnover in that analysis, which suggests that many employees who feel let down stay, but with less commitment and discretionary effort.
Individual Wellbeing and Stakeholder Impacts
For individuals, a decline in trust during a transformation is not merely an attitude shift. It typically comes bundled with uncertainty, and uncertainty is psychologically costly. Bordia et al. (2004) distinguished three types of change-related uncertainty: strategic (about where the organization is going), structural (about how work will be organized), and job-related (about one's own role, security, and future). In their study of a public-sector organization, job-related uncertainty was linked to psychological strain through a reduced sense of control. Management communication was associated with lower strategic uncertainty, while participation in decision-making was associated with lower structural and job-related uncertainty and with greater feelings of control. The implication is that employees need both information and influence, and that the two do different work.
Psychological contract breach adds an emotional layer. Robinson's (1996) longitudinal study of managers showed that employees with high initial trust in their employer were less likely to perceive breach and, when breach did occur, experienced a smaller subsequent decline in trust. Those who started out skeptical were more likely to interpret ambiguous events as breaches. This has an uncomfortable implication for organizations in a long run of change: as trust declines, employees become more likely to read each new initiative as a betrayal, which further depresses trust. It also explains why the same announcement can land very differently in two divisions of the same company.
The effects reach beyond individual employees. Broken trust can spill into public reputation, as several of the cases below illustrate. Edelman's (2026) data also suggest that eroding trust is fraying the social fabric inside workplaces: comfort discussing societal issues with colleagues who hold different views has fallen since 2022, by nine points on discrimination and six points on climate change. Trust in co-workers with different values was 77% among employees with a positive transformation experience and 50% among those with a negative one. Workplaces are among the few settings where people with different views cooperate daily, and that capacity seems to weaken when change is handled poorly.
There is also an equity dimension. The burden of poorly managed transformation falls hardest on associates and individual contributors, who in both the Edelman and DDI data report the worst experiences and the lowest trust (Edelman, 2026; Rhyne et al., 2026). Uneven erosion deepens the divide between those who make decisions and those who live with them.
Evidence-Based Organizational Responses
The research on trust does not offer a single fix, but it does point to a coherent set of practices. They follow logically from the trustworthiness model: people trust leaders who demonstrate competence, show genuine concern for their interests, and act with integrity; they extend that trust more readily when processes are fair, when they understand what is happening, and when they can influence it; and once trust has been violated, the response to the violation matters as much as the violation itself. Table 1 summarizes five responses and the evidence behind them.
Table 1: Evidence-based responses for building and repairing trust during transformation
Response | Evidence Base | Key Practices | Illustrative Examples |
Candid, reasoned communication about change | Mayer et al. (1995); Bordia et al. (2004); Rhyne et al. (2026) | Explain the why and how of decisions; share bad news early; state what is not known | Airbnb's 2020 restructuring letter; Buffer's published salary formula |
Procedural justice in restructuring and layoffs | Brockner et al. (1994); Oreg et al. (2011); Zhao et al. (2007) | Clear criteria, respectful delivery, generous support, leaders owning mistakes | Stripe's 2022 layoff; Better.com's 2021 Zoom layoff as a contrast |
Transparent AI and data governance | Glikson and Woolley (2020); Kemp and Pendell (2026); University of Melbourne (2025) | State purpose, limits, and job implications; talk directly with employees about data; invest in skills | Klarna's reversal on AI customer service; Duolingo's AI-first memo |
Managers as trust translators | Dirks and Ferrin (2002); Legood et al. (2021); Rhyne et al. (2026) | Coach and develop managers; brief them before announcements; reward trust-building in promotion | ADP's emphasis on listening sessions and reskilling |
Voice systems and trust repair after violations | Kim et al. (2004); Gillespie and Dietz (2009); Morrison (2014) | Match the response to the violation; fix systems, not only messages; close the loop on concerns | Boeing's safety culture review; Wells Fargo's sales practices response; NHS Freedom to Speak Up |
Lead Change With Candor and Visible Reasoning
The trustworthiness model suggests that communication during change is less about persuasion than about evidence. Employees are continually asking three questions of their leaders: Do they know what they are doing (ability)? Do they care what happens to us (benevolence)? Are they telling us the truth (integrity)? Each communication either supplies evidence on those questions or withholds it (Mayer et al., 1995). DDI's first recommendation follows this logic directly: show people how decisions are made by explaining the reasoning, giving clear direction, protecting confidentiality, and making it safe to raise concerns (Rhyne et al., 2026). The DDI data identify ethical decision making and concern for team members' wellbeing as the manager behaviors most strongly tied to trust, followed by clear direction, confidentiality, inspiration, and encouraging people to speak up.
Bordia et al. (2004) add a useful refinement: communication primarily reduces strategic uncertainty, the uncertainty about where the organization is going. Job-related uncertainty, the kind that keeps people awake at night, responds more to participation and a sense of control. Leaders who pour effort into polished strategy narratives while leaving employees in the dark about their own futures are addressing the wrong uncertainty. Edelman's (2026) advice to frame transformation in terms of productivity and wellbeing rather than disruption reflects a similar point: people want to understand what the change means for them, not only for the share price.
Effective approaches:
Explain the decision logic: Share the problem being solved, the options considered, and the criteria used, so employees can judge the leaders' competence and integrity for themselves.
Deliver bad news early and directly: Delays and leaks signal that leaders are protecting themselves, which damages perceived benevolence.
Name the unknowns: Say clearly what has not yet been decided and when it will be, rather than projecting false certainty that later proves wrong.
Address the personal question: Tell people what the change means for their roles, skills, and security as soon as that is known.
Make consistency visible: Publish commitments and then report against them, so employees can see words and deeds line up over time.
Airbnb faced the collapse of global travel in the spring of 2020. In a note to employees, CEO Brian Chesky explained that the company expected 2020 revenue to be less than half of 2019's, that it had raised capital and cut costs but needed more fundamental changes, and that it would lay off nearly 1,900 of its roughly 7,500 employees (Chesky, 2020). The letter described how decisions had been made: the company first determined the strategy and the size and shape of each team, then evaluated individuals on critical skills relative to future business needs. It detailed the support package, including 14 weeks of base pay plus one week per year of tenure for U.S. employees, 12 months of COBRA health coverage, an opt-in public alumni talent directory, and a recruiting team dedicated to placing departing employees. The letter did not make the news less painful, but it offered evidence on each dimension of trustworthiness: a reasoned business case, a transparent process, and tangible concern for those leaving.
Buffer, the social media software company, has taken transparency further as a standing practice rather than a crisis response. Since 2013 it has published its salary formula and, later, the complete list of employee salaries, explaining that transparency builds trust and holds the company accountable (Buffer, n.d.). The formula combines a benchmark percentile for each role with a cost-of-living band, so any employee can see how their pay was calculated. Few organizations will go this far, but Buffer's experience illustrates a principle that applies broadly: when the logic behind consequential decisions is visible in advance, employees have less reason to assume the worst when a decision goes against them.
Common pitfall: Leaders often mistake volume for candor. Frequent town halls and polished videos can coexist with a complete absence of the information employees actually want. If people leave a briefing knowing more about the strategy than about what happens to them next month, trust is unlikely to improve, and the gap between official messaging and lived experience may make it worse.
Make the Process Fair, Especially When the Outcome Hurts
Perhaps the single most practically useful finding in this literature concerns procedural justice, the perceived fairness of the processes used to make decisions. Brockner et al. (1994) studied three groups: employees who had been laid off, survivors who remained, and people who knew they would soon lose their jobs. Across all three, when procedural justice was low, more negative outcomes produced more negative reactions. When procedural justice was relatively high, the negativity of the outcome was no longer related to people's reactions. The fairness of the process buffered the pain of the result. This finding is profoundly relevant to transformation. Leaders often cannot avoid delivering bad outcomes, but they have a great deal of control over the fairness of the process.
Oreg et al.'s (2011) review reaches a compatible conclusion: the way a change is implemented, including the participation, communication, and support offered, shapes reactions independently of the change's content. And Zhao et al.'s (2007) meta-analysis reminds us that the costs of perceived breach fall on job satisfaction, commitment, and citizenship among those who remain. A layoff handled badly is not just an event that affects those who leave. It is a demonstration, watched closely by everyone who stays, of how the organization treats people when it no longer needs them.
Effective approaches:
Publish the criteria: Explain how roles and people were selected, and apply the criteria consistently.
Deliver news personally and privately: Ensure each affected person hears from a manager, in a one-to-one conversation, with time for questions.
Own the decision at the top: Senior leaders should acknowledge their role, including any misjudgments that led to the need for cuts.
Provide meaningful support: Severance, health coverage, placement help, and immigration assistance signal benevolence in ways words cannot.
Attend to survivors: Explain what changes for those who remain and how workload will be handled, rather than assuming relief will produce commitment.
Stripe cut 14% of its staff in November 2022. CEO Patrick Collison stated that the founders had made the decision, that the company had overhired, and that leadership had made mistakes (Estrada, 2022). Departing employees received 14 weeks of severance, their 2022 annual bonus, payment for unused paid time off, six months of health care coverage or its cash equivalent, accelerated vesting for those past their one-year cliff, career support, and dedicated immigration assistance for visa holders. Each person was to have a one-to-one conversation with a Stripe manager, and departing staff were given alumni email addresses. Fortune's coverage held the announcement up as an example of handling layoffs well, largely because the leadership accepted responsibility rather than blaming external conditions alone.
Better.com offers the contrast. In December 2021, CEO Vishal Garg laid off around 900 employees, roughly 15% of the workforce, during a Zoom webinar, citing market conditions, efficiency, and performance (Al Jazeera, 2021). A recording of the call circulated widely and drew public condemnation. Garg subsequently apologized, acknowledging that the way he had communicated the news made a difficult situation worse and that he had failed to show appropriate respect and appreciation for the affected employees, while still standing by the decision itself. The episode is instructive precisely because the business rationale was not obviously unreasonable. What did the damage was the process: impersonal delivery, no private conversation, and a framing that left those affected wondering whether they were being labeled poor performers.
Common pitfall: Treating procedural justice as a matter of legal compliance. Meeting notice requirements and offering standard severance does not by itself convey fairness. Employees evaluate whether they were treated with dignity, whether the criteria made sense, and whether leaders shared the pain, and those judgments are formed in conversations, not in policy documents.
Govern AI and Data Transparently
Because AI implementation is now the most common form of workplace transformation (Edelman, 2026), it deserves its own response. The trust challenge is twofold. Employees must decide whether to trust the AI tools themselves, and they must decide whether to trust the leaders who are deploying them and whose intentions regarding jobs and data are often unclear. Research on human trust in AI shows that transparency about how a system works is especially important for cognitive trust in embedded and virtual AI, and that initial trust can fall after use if expectations are not met (Glikson & Woolley, 2020). The University of Melbourne (2025) study suggests that organizations are falling well short on the basics: fewer than half of workers had received AI training, only 40% reported workplace guidance on generative AI, and 57% admitted hiding their AI use and presenting AI-generated work as their own, a pattern that itself suggests a lack of trust in how AI use will be judged.
On data, the Gallup findings are instructive because they point toward specific levers. Employees who had discussed workplace data use with someone at their organization were more likely to trust leaders a great deal (26% versus 18% among those who had not), although they were also somewhat more likely to express concern, which suggests that conversation surfaces issues as well as resolving them (Kemp & Pendell, 2026). Employees were also more comfortable with data use that supported them than with data use that might judge them: 62% were comfortable with activity data being used to help managers support team effectiveness, compared with 53% for formal performance evaluations. Gallup's recommendations are straightforward: be transparent about purpose and use, state what the organization will not do with the data, equip managers to answer questions, and build engagement more broadly.
Effective approaches:
State the employment implications plainly: Say what AI will and will not mean for headcount, roles, and hiring, and update that statement as plans evolve.
Explain purpose and limits of data use: Describe what is collected, why, who can see it, and what it will never be used for.
Invest in capability before expectation: Provide training and guidance before making AI use a performance requirement.
Keep humans visible: Where AI affects customers or employees, make clear when and how a person can intervene.
Pilot and report honestly: Share what is working and what is not, including reversals, rather than presenting AI as an unqualified success.
Klarna shows how bold AI claims can outrun reality. The company said in 2024 that its AI assistant was doing the work of 700 customer service agents, it reduced its headcount by more than 1,000 through AI-driven efficiencies, and its CEO told Bloomberg in early 2025 that he believed AI could already do all the jobs humans do (Ivanova, 2025). By May 2025, the CEO acknowledged that cost had been too dominant a factor in its AI decisions and had led to lower quality, and the company began recruiting human agents so customers could always reach a person, while a spokesperson maintained that Klarna remained AI-first. The reversal shows the trust risk of announcing AI's capabilities before they have been demonstrated: employees are left to wonder which statements to believe.
Duolingo encountered a different version of the same problem. In April 2025, CEO Luis von Ahn shared a memo describing an AI-first strategy that included gradually stopping the use of contractors for work AI could handle and allowing headcount growth only where teams could not automate more of their work (Coacci, 2025). The memo provoked a public backlash and fears of mass layoffs. Months later, von Ahn said the fault was his own and that he had not given enough context, adding that the company had never laid off full-time employees and did not plan to replace staff with AI. The episode illustrates how an AI announcement framed primarily in terms of efficiency invites the worst interpretation, and how much harder it is to supply context after the fact than before.
Common pitfall: Communicating about AI only in the language of productivity. Employees hear efficiency as a euphemism for job cuts unless leaders explicitly address what happens to people. Silence on the employment question is not neutral; in a climate where three in four workers already fear job loss (Edelman, 2026), it will be filled with assumptions.
Equip Managers to Translate and Sustain Trust
If direct leaders are a particularly important referent of trust (Dirks & Ferrin, 2002), then managers are where trust during transformation is won or lost. They are the people who explain the change, absorb the questions, and decide how to implement it on the ground. Yet the DDI data show that managers themselves are among the least trusting groups in the organization: only 31% trust their own manager and 27% trust senior leaders (Rhyne et al., 2026). Organizations are effectively asking a skeptical middle layer to sell change they may not believe in. The DDI findings on development are therefore striking. Leaders whose managers support their development are 4.8 times more likely to trust them, and those who believe their managers coach them effectively are 9 times more likely to trust them. Trust in managers is also associated with finding work meaningful (1.9 times), feeling energized (2.1 times), and feeling prepared for change (2 times).
These associations do not establish causation, since employees who trust their managers may simply describe them more charitably. They are, however, consistent with meta-analytic evidence that ethical and servant leadership predict trust, and that affective trust predicts performance and citizenship (Legood et al., 2021). The DDI authors also note that proximity alone does not build trust: in their data, trust in managers was 44% among remote workers, 36% among hybrid workers, and 31% among those fully in person (Rhyne et al., 2026). What matters is what managers do, not how often they are seen.
Effective approaches:
Brief managers first: Give managers the reasoning, the facts, and the likely questions before employees hear about a change, and let them shape how it is rolled out.
Develop coaching capability: Treat coaching as a core managerial skill, with training, practice, and feedback, rather than an optional extra.
Reward trust-building: Include integrity, coaching, and team trust in talent reviews and promotion decisions alongside results.
Equip managers on sensitive topics: Give them visibility into AI and data practices so they can answer employees' questions honestly.
Support the supporters: Recognize that managers absorb a great deal of emotional labor during change, and provide them with peer forums and their own coaching.
ADP, the payroll and human capital management firm, has made trust a theme of its own leadership messaging. In an October 2026 interview, CEO Maria Black argued that leaders should start by asking whether their employees trust them, stay close to employees through visits and listening sessions, and invest in reskilling people to work with AI tools (Gerut, 2026). The company's chief talent officer has also promoted practical tips on building trust, linked to ADP research associating trust in senior leadership with motivation and commitment (ADP, n.d.). The emphasis on direct listening and reskilling aligns with the DDI evidence on development, although the public sources do not describe how these practices are implemented internally or with what measured effect.
Common pitfall: Using managers as a distribution channel rather than as partners. When managers receive change announcements at the same time as their teams, with talking points instead of reasoning, they cannot answer the questions that matter, and their own credibility suffers. DDI's warning that managers will hedge or pass decisions along without conviction describes the predictable result (Rhyne et al., 2026).
Build Voice and Repair Trust After Violations
No organization going through repeated transformation will avoid every trust violation. Plans change, promises turn out to be unkeepable, and sometimes leaders make genuine mistakes. The research on trust repair offers specific guidance on how to respond. Kim et al. (2004) found in two experiments that the effectiveness of an apology versus a denial depended on the nature of the violation. For violations that seemed to reflect a lack of competence, apologizing repaired trust more effectively, because people believe competence can improve and a single failure is not diagnostic. For violations that seemed to reflect a lack of integrity, denial worked better when it was true, because people treat a single integrity failure as strongly diagnostic of character. When evidence later showed guilt, apology outperformed denial. The practical lesson is that leaders should diagnose what kind of violation employees believe has occurred, and that denial is only an option when it is honest.
For organization-wide failures, individual apologies are not enough. Gillespie and Dietz (2009) argue that an organization's trustworthiness is signaled by multiple components at once, including leadership, culture, strategy, structures, policies, and external reputation, and that repair requires addressing the systemic causes of failure, not only the communication. They propose a staged process: an immediate response that acknowledges the incident and expresses regret, a diagnosis of what went wrong, reforms that address the causes, and an evaluation of whether the reforms are working. Dirks et al. (2009) similarly emphasize that repair involves restoring both positive expectations and the structures that make future violations less likely.
Voice systems are the bridge between prevention and repair. Edmondson (1999) showed that psychological safety, a shared belief that a team is safe for interpersonal risk taking, supports the learning behaviors through which teams surface and correct problems. A meta-analysis of 136 samples confirmed psychological safety's links with task performance and citizenship behavior (Frazier et al., 2017). Morrison (2014) emphasizes that voice depends on both safety and efficacy, meaning employees must believe that speaking up will not harm them and that it will actually change something. Organizations that solicit input during transformation and then ignore it can do more harm than those that never ask.
Effective approaches:
Match the response to the violation: Apologize sincerely for competence failures; for integrity failures, be truthful, and if wrongdoing occurred, acknowledge it fully.
Fix the system, not just the story: Identify the incentives, structures, and policies that produced the failure and change them visibly.
Impose real accountability: Consequences for senior leaders signal that integrity applies at the top.
Create independent channels for concerns: Separate the investigation of concerns from the reporting lines that control people's pay and promotion.
Close the loop: Tell people what happened as a result of their concerns, even when the answer is that nothing could be changed.
Boeing provides a cautionary picture of what happens when the voice system itself lacks trust. An expert panel convened under federal legislation to review Boeing's safety management and organizational delegation practices reported in February 2024 with 27 findings and 53 recommendations (Federal Aviation Administration, 2024). The panel observed a disconnect between senior management and other members of the organization on safety culture. It found that managers who controlled employees' reviews, pay, and promotions could also investigate safety reports within their own reporting chains, leaving some employees hesitant to report for fear of retaliation, and that many employees distrusted the anonymity of the internal Speak Up program and often did not learn the outcome of their reports. The recommendations included more autonomous investigation processes and better feedback to reporters, which correspond closely to the safety and efficacy conditions Morrison (2014) describes.
Wells Fargo illustrates the systemic approach to repair advocated by Gillespie and Dietz (2009), at least on paper. After the retail banking sales practices scandal, an independent board investigation released in April 2017 identified as root causes the sales culture and performance management system in the community bank, a decentralized structure that gave its leadership too much autonomy, resistance to escalating problems, and a slow response from the CEO (Wells Fargo, 2017). The board announced clawbacks and forfeitures from former executives, including approximately $69 million from former CEO John Stumpf and $67 million from former community banking head Carrie Tolstedt, separated the chair and CEO roles, centralized risk and human resources functions, and eliminated product sales goals from retail compensation. Those reforms targeted the incentive system that had pressured employees in the first place, though restoring trust fully was a longer project than any single announcement.
The NHS in England offers an example of building voice infrastructure at scale after failure. Following the 2015 Freedom to Speak Up review, the National Guardian's Office was established in 2016 to lead and support a network of Freedom to Speak Up Guardians in health care organizations (NHS England, 2024). By May 2024 the network included 1,231 guardians, who handled 32,167 cases in 2023/24, a 27% increase on the prior year. The same report candidly noted the limits of progress: in the 2023 NHS Staff Survey, only 62.3% of staff felt safe speaking up about anything that concerned them, and only 50.1% believed their organization would address such concerns. That second figure is a reminder that voice systems build trust only when people see their concerns acted on.
Common pitfall: Treating trust repair as a communications exercise. Apologies and new values statements are necessary, but employees judge sincerity by whether underlying systems change. Repair that stops at the message can deepen cynicism, because it confirms the suspicion that leaders care more about appearances than about the problem.
Building Long-Term Trust Capability
Measure Trust Where Decisions Become Action
Most organizations measure engagement. Far fewer measure trust directly, and fewer still measure it by referent and by level. Yet the evidence reviewed here suggests that aggregate measures can be badly misleading. DDI's data show that executives see a healthier picture than their organizations' middle layers, and Edelman's data show that the experience of transformation differs sharply by level (Edelman, 2026; Rhyne et al., 2026). DDI's recommendation to measure trust at every level of leadership, not just among executives, is simple but rarely practiced.
A practical measurement approach would assess trust separately in the direct manager, senior leaders, and co-workers, and would include items on the three dimensions of trustworthiness so that leaders can see whether the problem lies in perceived ability, benevolence, or integrity (Colquitt et al., 2007; Mayer et al., 1995). It would ask employees about their experience of specific changes, not only about change in general, and it would track whether employees believe their concerns are addressed, the efficacy side of voice (Morrison, 2014). Results should be shared openly and discussed with employees, which is itself a demonstration of integrity. Gallup's finding that a conversation about data increases trust suggests that the act of discussing these issues is part of the intervention, not just a precursor to it (Kemp & Pendell, 2026).
Build Trust Before It Is Needed
DDI's third recommendation captures an insight that runs through the research: trust develops through repeated experiences over time, and it must exist before disruption hits (Rhyne et al., 2026). Robinson's (1996) finding that high initial trust reduced both the perception of breach and the decline in trust after breach suggests that trust functions as a kind of reserve. Organizations that have built trust through consistent, fair behavior in ordinary times can draw on it when they must ask employees to accept uncertainty. Those that have not are asking for credit they do not have.
Building that reserve is mostly a matter of routine behavior: keeping small promises, explaining decisions even when not obliged to, consulting employees before changes rather than after, and being honest about failures, such as what did not work in a pilot or why a plan changed. A recent review of a quarter-century of workplace trust research anticipates that the next wave of scholarship will focus on how organizational change and societal disruption affect trust (Dirks & de Jong, 2022). The practical corollary is that trust can no longer be treated as a stable background condition; it has to be deliberately maintained.
Leadership selection and development are central here. If the behaviors most associated with trust are ethical decision making, concern for wellbeing, clear direction, confidentiality, and encouraging voice (Rhyne et al., 2026), and if servant and ethical leadership predict trust beyond more traditional leadership styles (Legood et al., 2021), then organizations should select and promote for those behaviors explicitly. Promotion systems that reward delivery regardless of how it was achieved will gradually populate the management ranks with people whose results come at the expense of trust.
Institutionalize Fairness and Voice in Change Governance
The final capability is structural. In many organizations, decisions about transformation are made in a small circle and communicated outward. The evidence on procedural justice and participation suggests that this model is costly (Bordia et al., 2004; Brockner et al., 1994; Oreg et al., 2011). An alternative is to build fairness and voice into the governance of change itself: standard requirements that major initiatives include an assessment of their effect on employees, documented criteria for decisions affecting jobs, structured consultation with employees or their representatives before decisions are final, and independent channels for raising concerns during implementation.
For AI specifically, this might mean a published set of principles describing how AI will be used, what employee data it will draw on, how decisions about roles will be made, and what support will be offered to those whose work changes. The University of Melbourne (2025) finding that most workplaces lack guidance on generative AI suggests considerable room for improvement. Gallup's evidence that employees are more comfortable with data used to support them than to evaluate them offers a starting point for such principles (Kemp & Pendell, 2026).
Boeing's experience suggests that voice channels must be structurally independent of the chain of command that controls careers (Federal Aviation Administration, 2024), and the NHS experience suggests that voice infrastructure can be built at scale but earns credibility only through visible responsiveness (NHS England, 2024). Because trust is signaled by every part of an organization at once (Gillespie & Dietz, 2009), governance that aligns policies, incentives, and leadership behavior is more likely to hold than any single initiative.
Conclusion
The evidence assembled here suggests that trust in leaders, managers, and colleagues is under real strain, and that the pace and character of transformation, especially AI implementation, are part of the reason. The more important message, though, is that the strain is not inevitable. Employees who experienced transformation positively reported trust in their CEO 33 points higher than those who experienced it negatively (Edelman, 2026). How change is led appears to matter enormously.
Decades of research offer a clear account of what leaders can do. People trust those who demonstrate ability, benevolence, and integrity, and they judge those qualities from behavior, especially under pressure (Mayer et al., 1995; Colquitt et al., 2007). Fair processes can blunt the harm of painful outcomes (Brockner et al., 1994). Communication and participation address different kinds of uncertainty (Bordia et al., 2004). Direct managers are a critical referent of trust and need to be equipped and supported (Dirks & Ferrin, 2002; Rhyne et al., 2026). And when trust is broken, the response must fit the violation and reach the systems that produced it (Gillespie & Dietz, 2009; Kim et al., 2004). None of this is mysterious. It is, however, demanding, because it requires leaders to treat trust as something earned continuously rather than something conferred by position.
Lead with reasons, not just announcements: Explain how decisions were made and what they mean for people's roles, and name what is not yet known.
Protect the process when outcomes hurt: Use clear criteria, personal delivery, meaningful support, and senior ownership in every restructuring.
Govern AI and data in the open: State the employment implications and data limits of AI plainly, and invest in skills before setting expectations.
Make managers partners in change: Brief them first, coach them well, and reward the trust they build alongside the results they deliver.
Build voice that works: Create independent channels for concerns, close the loop on what happens, and treat silence as a warning sign.
Repair at the level of the system: After a violation, match the response to the breach and change the incentives and structures that caused it.
Measure trust where it matters most: Track trust by referent and level, especially among the people who turn decisions into action.
Research Infographic
References
ADP. (n.d.). Newsroom spotlight: How trust can fuel business success. ADP Media Center.
Al Jazeera. (2021, December 8). 'Better' boss apologises for mass-firing 900 workers on Zoom call. Al Jazeera.
Bordia, P., Hobman, E., Jones, E., Gallois, C., & Callan, V. J. (2004). Uncertainty during organizational change: Types, consequences, and management strategies. Journal of Business and Psychology, 18(4), 507–532.
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.
Buffer. (n.d.). Transparent salaries. Buffer.
Chesky, B. (2020, May 5). A message from co-founder and CEO Brian Chesky. Airbnb Newsroom.
Coacci, J. (2025, August 18). Duolingo CEO admits his controversial AI memo "did not give enough context" and insists the company never laid off full-time employees. Fortune.
Colquitt, J. A., Scott, B. A., & LePine, J. A. (2007). Trust, trustworthiness, and trust propensity: A meta-analytic test of their unique relationships with risk taking and job performance. Journal of Applied Psychology, 92(4), 909–927.
Dirks, K. T., & de Jong, B. (2022). Trust within the workplace: A review of two waves of research and a glimpse of the third. Annual Review of Organizational Psychology and Organizational Behavior, 9, 247–276.
Dirks, K. T., & Ferrin, D. L. (2002). Trust in leadership: Meta-analytic findings and implications for research and practice. Journal of Applied Psychology, 87(4), 611–628.
Dirks, K. T., Lewicki, R. J., & Zaheer, A. (2009). Repairing relationships within and between organizations: Building a conceptual foundation. Academy of Management Review, 34(1), 68–84.
Edelman, R. (2026, October 6). Trust at work: Transformation triggering trust demise. Edelman.
Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383.
Estrada, S. (2022, November 7). As layoffs mount, one tech CEO figured out how to do them right. Fortune.
Federal Aviation Administration. (2024). Section 103 organization designation authorizations (ODA) for transport airplanes: Expert panel review report. Federal Aviation Administration.
Frazier, M. L., Fainshmidt, S., Klinger, R. L., Pezeshkan, A., & Vracheva, V. (2017). Psychological safety: A meta-analytic review and extension. Personnel Psychology, 70(1), 113–165.
Gerut, A. (2026, October 15). ADP CEO Maria Black says employees who trust their CEO and senior leaders are 41 times more likely to be engaged at work. Fortune.
Gillespie, N., & Dietz, G. (2009). Trust repair after an organization-level failure. Academy of Management Review, 34(1), 127–145.
Glikson, E., & Woolley, A. W. (2020). Human trust in artificial intelligence: Review of empirical research. Academy of Management Annals, 14(2), 627–660.
Ivanova, I. (2025, May 9). As Klarna flips from AI-first to hiring people again, a new landmark survey reveals most AI projects fail to deliver. Fortune.
Kemp, A., & Pendell, R. (2026, September 24). Employee data use: One in five trust leaders a great deal. Gallup.
Kim, P. H., Ferrin, D. L., Cooper, C. D., & Dirks, K. T. (2004). Removing the shadow of suspicion: The effects of apology versus denial for repairing competence- versus integrity-based trust violations. Journal of Applied Psychology, 89(1), 104–118.
Legood, A., van der Werff, L., Lee, A., & Den Hartog, D. (2021). A meta-analysis of the role of trust in the leadership-performance relationship. European Journal of Work and Organizational Psychology, 30(1), 1–22.
Mayer, R. C., Davis, J. H., & Schoorman, F. D. (1995). An integrative model of organizational trust. Academy of Management Review, 20(3), 709–734.
Morrison, E. W. (2014). Employee voice and silence. Annual Review of Organizational Psychology and Organizational Behavior, 1, 173–197.
NHS England. (2024). National Guardian's Office update and priorities for 2024/25. NHS England.
Oreg, S., Vakola, M., & Armenakis, A. (2011). Change recipients' reactions to organizational change: A 60-year review of quantitative studies. Journal of Applied Behavioral Science, 47(4), 461–524.
Rhyne, R., Abraham, E., Byham, T. M., Creedy, V., & Tamanini, K. (2026, October 14). Trust is lowest among those carrying out leadership decisions (Global Leadership Forecast 2027). DDI.
Robinson, S. L. (1996). Trust and breach of the psychological contract. Administrative Science Quarterly, 41(4), 574–599.
Rousseau, D. M., Sitkin, S. B., Burt, R. S., & Camerer, C. (1998). Not so different after all: A cross-discipline view of trust. Academy of Management Review, 23(3), 393–404.
University of Melbourne. (2025, May 23). Global study reveals trust of AI remains a critical challenge reflecting tension between benefits and risks [Media release]. University of Melbourne.
Wells Fargo. (2017, April 10). Wells Fargo board releases findings of independent investigation of retail banking sales practices and related matters [Press release]. Wells Fargo.
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). Trust in Freefall: Why Relentless Transformation Is Eroding Confidence in Leaders, and How Organizations Can Rebuild It. Human Capital Leadership Review, 40(3). doi.org/10.70175/hclreview.2020.40.3.4
