AI transformation is not merely a technology project involving the installation of a new system. For employees, AI is both a tool that can improve productivity and a presence that may threaten their roles and expertise. The success or failure of AI adoption depends less on technological performance than on leadership that understands anxiety and creates space for trust and learning.

***

Companies are viewing generative AI as a tool capable of achieving productivity gains, cost reductions, and business innovation at the same time. As the potential applications of AI have been demonstrated across a wide range of tasks?including report writing, information retrieval, customer service, data analysis, programming, and design?the focus of corporate attention has shifted from whether AI should be adopted to how quickly and broadly it should be used. Senior executives announce enterprise-wide AI strategies, introduce new platforms, and provide training programs for employees. However, installing the latest systems and distributing user accounts does not mean that an organization’s AI transformation is complete.

Technology can be installed within a short period, but people’s ways of working do not change so quickly. Employees must abandon procedures with which they are familiar and learn new tools, while also reconsidering whether the experience and expertise they have accumulated will remain valuable in the future. It is also unclear whether AI will serve as a tool that supports their work, a competitor that replaces part of it, or even a force that threatens their jobs. In this environment, AI adoption becomes not simply a matter of learning functions but a process of reinterpreting the meaning of work and one’s own position within the organization.

Many companies overlook this point. Senior executives expect that employees will naturally accept new tools once the capabilities of AI and the expected returns on investment are explained. They emphasize how much productivity can improve, how much processing time can be reduced, and what competitive advantages can be gained over rivals. Employees, however, are interested in somewhat different questions. They want to know how the time saved after work becomes faster will be used, whether performance expectations will also rise when AI improves output, and what roles they will be expected to assume if the work they currently perform becomes automated.

When a company emphasizes only the speed of adoption without answering these questions, employees may perceive AI not as an opportunity for innovation but as an uncertain signal the organization is sending about their future. They may formally use the new system while continuing to rely on existing methods in actual practice, using only the simplest functions or withholding their trial-and-error experiences from colleagues. The technology may have entered the organization, but the way work is done remains unchanged.

The essence of AI transformation lies not in deploying technology but in redesigning the relationship between people and technology. Organizations need agreement on which tasks should be entrusted to AI and which judgments should remain with people, who should review and take responsibility for AI-generated results, and how the time saved should be redirected toward value creation. When employees are excluded from this process, AI is perceived as a tool of control imposed from above. When employees participate in designing and testing the change, however, AI can become a tool that expands their capabilities.

The leaders responsible for AI adoption therefore need more than an understanding of technology. They must observe what emotions and questions the change creates for employees, discuss issues for which no final answers yet exist, and create an environment in which new ways of working can be tested safely. Competitiveness in the AI era will emerge not from companies that purchase the best systems first, but from those that enable their people to trust and use those systems.

Passive attitudes shown by employees during AI adoption are often explained as resistance to change. They may be interpreted as an unwillingness to learn new technology or a desire to remain comfortable with existing methods. Yet when resistance is viewed solely as a problem of individual motivation or adaptability, organizations overlook important causes they themselves need to address. Behind employee behavior lie three different forms of anxiety: job loss, the erosion of expertise, and the burden of learning.

The first is anxiety about job loss. The more strongly a company links the purpose of AI adoption to workforce reduction or cost cutting, the more likely employees are to view the new technology as something that will take their work away. They may believe that actively using AI will automate their tasks and, as a result, reduce the need for their own roles. In such a situation, encouraging employees to use AI can sound like asking them to discover for themselves how they might be replaced.

Anxiety becomes even greater when the organization has reduced its workforce during previous automation initiatives or mentions the possibility of restructuring while introducing AI. Even if management says that “AI will not replace people,” employees are unlikely to trust the statement unless there are concrete plans for role transitions and retraining. What matters is not vague reassurance but a clear explanation of which tasks will change, which capabilities will become necessary, and how the organization will support employees through the transition.

The second is anxiety about the loss of expertise. The knowledge and experience employees have accumulated over many years mean more than simple job competence. They form the basis on which employees confirm their own value within the organization and receive recognition from colleagues. When AI can draft a report in seconds, propose ideas, or carry out part of an analysis once performed by specialists, employees may feel that the value of the expertise they have built is declining.

This anxiety does not arise simply because AI performs well. It becomes more intense when employees do not know what they will be evaluated for in the future. In the past, the ability to locate accurate information and produce documents may have been an important performance criterion. In the AI era, however, the ability to formulate questions, verify outputs, connect multiple sources of information, and exercise judgment may become more important. Unless the organization explains these new evaluation criteria, employees may feel that their existing strengths are disappearing while only the skills they lack are being highlighted.

The third is anxiety about the burden of learning. Learning new technology requires time and concentration. Yet many organizations add AI training and usage expectations while leaving existing workloads unchanged. Employees must handle their regular responsibilities while learning new tools, identifying appropriate uses on their own, and reviewing outputs to prevent errors. Contrary to the claim that AI will reduce work, employees may initially find that it creates even more tasks.

Employees also differ in their experience with and confidence in digital technology. Some quickly experiment with new tools, while others worry that using them incorrectly will make them appear incompetent. Some remain silent because they fear that asking questions may create the impression that they do not understand even the basics. When organizations highlight only the success stories of employees who adapt quickly, those who learn more slowly may become even more discouraged.

As generative AI begins to perform cognitive, creative, and interpersonal tasks, employees may feel that their competence, autonomy, and sense of belonging within the organization are under threat. This shows that resistance to AI is not merely a fear of technology but a problem connected to professional identity.

The leader’s role is not to dismiss employee anxiety as an irrational emotion. Leaders must ask specifically what employees are most concerned about and establish systems that respond to those concerns. When role changes are expected, they should provide retraining and mobility pathways. When learning is required, employees should be given opportunities to practice during working hours. When the value of expertise is changing, leaders must explain clearly which capabilities will become more important. Understanding anxiety does not slow change; it removes the causes of resistance.

One of the most dangerous problems in AI adoption is that senior executives and employees may be looking at entirely different realities. Management may judge that adoption is proceeding smoothly based on formal reports, training participation rates, and system login statistics. Yet such figures do not reveal how effectively employees are using AI in their actual work, what difficulties they are experiencing, or how much they trust the quality of the results.

A survey of 1,400 employees in the United States found that 76 percent of senior executives believed employees were enthusiastic about their organization’s AI adoption, while only 31 percent of frontline employees actually expressed such enthusiasm. The gap between management’s perception and employees’ experience was more than twofold.

This gap becomes even wider in organizations where employees find it difficult to speak honestly. Once a leader has declared AI adoption to be a core strategy, it is not easy for employees to raise concerns. If they say that AI is not providing much help or that it takes more time than existing methods, they may fear being labeled as resistant to change. Employees may therefore respond positively in public while avoiding actual use in their day-to-day work.

The problem also worsens when leaders demand only productivity metrics. When employees are pressured to produce more using AI, they may use AI-generated content without sufficient review or hide the fact that they used AI and submit only the final result. Because errors and limitations are not shared, the organization repeats the same mistakes. AI usage experiences remain with individuals, and organization-wide learning does not occur.

Leadership without empathy mistakes employee silence for agreement. Because no objections are expressed, leaders assume that adoption is progressing well. Because training sessions end without disruption, they assume everyone has understood. Silence, however, may not be agreement. It may be a defensive behavior chosen to avoid risk. An organization in which questions and concerns have disappeared is not necessarily an organization without resistance; it may be one in which resistance has simply become invisible.

Empathetic leaders take a different approach. They do not assume that employees avoid AI because they lack motivation. Instead, they examine both the working environment and the emotions surrounding use. They ask which functions have actually been helpful, where more time has been required, which errors have caused the greatest anxiety, and which questions have been difficult to raise openly. What matters is not simply listening but incorporating the identified problems into policies and support systems.

For example, if employees are spending significant time verifying AI responses, leaders should not simply demand higher usage rates. They should establish validation standards and define the scope of responsibility. If employees are confused about whether sensitive data may be entered into an AI system, the organization should provide clear security guidelines. If employees are worried about how evaluation and compensation will change once workloads are reduced, performance standards should be redesigned. Empathy begins with understanding another person’s feelings, but it becomes leadership only when it leads to changes in systems and behavior.

To reduce the AI adoption gap, senior leaders must be able to hear uncomfortable truths before positive news. Employees should be able to say, “This tool is not helping yet,” and leaders should treat that statement not as opposition to change but as information necessary for improvement. The more quickly frontline difficulties and failures are shared, the faster the organization can adapt AI to actual work.

The value of AI cannot be fully discovered simply by reading a fixed set of instructions. The same tool can be used differently depending on the role and situation. Employees find the applications best suited to their work by trying different questions, comparing results, and learning from the experiences of colleagues. AI adoption is not an event completed through a single training session. It is a learning process in which organizational members experiment together and accumulate knowledge about how the technology can be used.

Experimentation, however, involves the possibility of failure. Employees may receive inaccurate answers, enter inappropriate prompts, or spend more time than they would using existing methods. If they believe that revealing such trial and error will lead to criticism or disadvantage in performance evaluations, experimentation will decline. They will repeatedly use only the safest functions or follow examples already verified by others. The organization may have adopted AI, but it will fail to create new ways of using it.

Connection and trust provide the foundation that allows employees to take these risks. When employees believe that their opinions will be respected, that disclosing mistakes will not cause them to be labeled incompetent, and that questions can lead to mutual assistance, they become more willing to experiment with new tools. Psychological safety does not mean accepting every opinion or lowering performance standards. It means creating conditions in which problems, errors, and opposing views can be expressed rather than hidden.

Employees who feel understood are more likely to share their experiences honestly. They tell colleagues which types of questions produced better results, which tasks were not suitable for AI, and what should be checked when reviewing AI outputs. Individual trial and error is transformed into organizational knowledge, reducing the cost of repeating the same mistakes. Collaboration also becomes more effective.

Leaders should avoid turning AI use into an individual competition. If they emphasize only those who produce the most output or learn the technology most quickly, employees may hide their know-how and conceal failure. By contrast, when organizations recognize not only successful use cases but also the sharing of failures and limitations, they learn collectively.

It can also be helpful to create small learning communities at the working level. Employees performing similar tasks can meet regularly to exchange use cases and jointly review the quality and risks of AI-generated results. The important point is not merely to list successful prompts. Participants should share the situation in which AI was used, the additional judgment contributed by a person, and who took responsibility for the final result.

It is also important for leaders to demonstrate that they themselves are learning. When leaders act as though they already know every answer about AI, employees become reluctant to reveal what they do not know. When leaders acknowledge that they too are learning the technology, have encountered poor results, and remain uncertain about some judgments, the barrier to asking questions and experimenting becomes lower. Acknowledging uncertainty does not weaken a leader’s authority. It allows the organization to see reality more accurately.

In organizations where employees feel cared for and respected, they are more likely to take creative risks, collaborate, and propose ideas. AI usage also expands on this relational foundation. The potential of technology depends not on the number of functions it offers but on how freely people can connect and share their experiences.

In traditional change management, leaders were described as people who set direction and persuade others to follow. They established clear goals, schedules, and execution methods, and then required employees to comply. AI transformation is different because its results cannot be predicted fully in advance. The technology continues to develop, and role-specific applications are still being established. A method that is effective today may change within a few months, while the tasks that become automated and the capabilities that grow more important will continue to be adjusted.

In this environment, it is dangerous for leaders to behave as though they already possess every answer. They may become unable to acknowledge problems emerging in the field and may regard revising an original plan as a failure. Employees, in turn, begin reporting what they think leaders want to hear, causing the organization to drift further away from reality.

What leaders need in the AI era is not perfect certainty but the humility required to lead learning. They should provide clear direction regarding why the organization is adopting AI, while discovering specific ways of using it together with employees. Rather than stopping at the abstract goal of improving productivity, they should discuss which forms of wasted effort should be reduced and what valuable activities should receive the time that is saved.

Leaders must also explain who will benefit from the gains created by AI. If efficiency improves but employees are simply given more work, AI will be perceived as a tool that intensifies the burden. Employees will participate voluntarily only when they can use the time saved for customer communication, creative planning, professional development, rest, and recovery, and when they believe that the organization and its people will share the value created by technology.

Empathetic leadership is different from unconditional reassurance or optimistic language. It does not mean accepting every employee demand or slowing the pace of change. It is an approach that accurately identifies employee anxiety, transparently explains the purpose and impact of change, and provides the learning resources and choices employees need. When leaders listen but change nothing, employees interpret empathy as a performative procedure.

Substantive empathy appears in several concrete behaviors. Leaders should honestly explain the purpose of AI adoption and the expected changes in roles to the greatest extent possible. They should create channels through which employees can raise questions and concerns and then explain how those views have influenced decision-making. Learning should not be left entirely to the individual; employees should receive working time and educational resources. Organizations should evaluate not only AI-based results but also the processes of verification, responsibility, and collaboration.

Above all, leaders must recognize that employees are not the objects of AI transformation but its agents. Frontline employees understand better than anyone which tasks are repetitive, where customers experience inconvenience, and which decisions require human experience. An AI strategy that excludes their knowledge may appear technically sophisticated but is unlikely to fit actual work. Employee participation is not merely a procedure for obtaining consent. It is a core condition for designing a better system.

Companies cannot directly control the performance of AI. Competing firms can use similar models and services, and technological gaps can narrow quickly. By contrast, the trust, collaborative practices, and learning culture accumulated within an organization are difficult to replicate. Even when using the same AI, organizations in which employees freely share experiences, propose new applications, and report problems early will create value more quickly.

The success or failure of AI adoption is not determined by whether a company possesses the largest number of functions. What matters is whether employees feel safe using the technology, whether they believe their future is being respected throughout the change, and whether they can learn together through mistakes and questions. Technology provides possibilities, but people turn those possibilities into performance.

Empathy is not a soft decoration added to AI innovation. It is the foundation that transforms anxiety into participation, silence into dialogue, and individual trial and error into organizational learning. Companies that treat human emotion as an obstacle to technological adoption may achieve formal usage rates, but they will struggle to create sustainable innovation. Companies that place people at the center of change can develop AI from a simple automation tool into a force that expands human judgment and creativity.

An outstanding leader in the AI era is not someone who possesses every answer about the future. It is someone who acknowledges what remains uncertain, listens to employees’ fears and expectations, and enables the organization to search for answers together. AI transformation is not a competition over how quickly technology can change people. It is a competition over who can first build an organization in which people can grow alongside technology.

Harvard Business Review, April 2026, Empathetic Leadership Can Make or Break AI Adoption.