Why Employees Resist AI—and How Companies Can Win Them Over

Some technologies face resistance not because they don’t work well, but because they could change the professional roles of the people who use them. Automation and AI are a sharp example: They can improve performance while making employees feel less expert or visible in their jobs.

That, say Das Narayandas and Shunyuan Zhang, professors at Harvard Business School, is the distinctive problem of self-disruptive technologies.

These concerns help explain a striking forecast. Despite the promise of artificial intelligence to ease workloads, at least 30% of generative AI projects will be abandoned—often, the researchers argue, not because the tools fall short, but because they are quietly rejected by the very employees they were designed to help, who see them as a threat to their professional roles.

In studying B2B sales, the researchers argue that changing the framing can help with adoption: New technologies gain traction not just when they improve performance, but when businesses make clear that the tools support—rather than threaten—the identity, expertise, and influence of the people who approve and use them.

Previous analyses of resistance to these tools have focused on company-level explanations, including institutional pressures or lack of organizational readiness, says Narayandas, the Edsel Bryant Ford Professor of Business Administration. “They ignore a very important gap: the motivations of individuals resulting from a crisis they are facing around their own identity,” he says.

While Narayandas has noticed for decades that teams shun automation that takes over their daily tasks, people may see generative AI as particularly disruptive to their sense of identity because of its ability to mimic higher-order human functions, including data analysis, decision-making, and writing work, he says.

“The extent to which people feel threatened and have a negative reaction to technology has only gotten worse” as AI has advanced, agrees Zhang, the F. Warren McFarlan Associate Professor of Business Administration.

Narayandas and Zhang cowrote the working paper “Selling Self-Disruptive Technologies: Identity-Compatible Advantage and the Role-Level Microfoundations of Automation Adoption,” released in February.

An AI-inspired identity crisis

In analyzing why technology projects flop, the authors identified three ways AI could threaten the identity and credibility of managers and other users of the technology:

• Role compression. As judgment and expertise are automated along with daily tasks, employees may find themselves with less to do overall, and the duties they do have may feel lower in status. “You still have a job, but the machine is doing all the interesting things,” Narayandas says. “We used to joke that as an MBA at a consulting company, you begin by learning how to staple before earning your way to making decisions. It might be AI is taking us in the opposite direction, where further on in your career you are back to stapling presentations, and the high-status tasks are taken over by machines.”

• Control shift. As decisions are delegated to algorithms, supervisors may have less discretion over the choices that once defined their expertise. “It's like driving a Tesla, where your hands are on the wheel, but you're not the one turning it,” Narayandas says. What thins out is the user's sense of authorship: The judgment that made the role feel like theirs now sits with the system, even though the title and the seat remain.

• Span erosion. As more people in the company begin to rely on AI for answers, managers’ influence over people, budgets, and processes may decrease, leaving them with a weaker sense of status overall. “You’re now royalty without a kingdom,” says Narayandas. “You still have your title, but the traditional sense of power within a job disappears.”

Faced with this triple threat to their identity, technology approvers and users have three choices in how to react, the authors say: They can refuse to comply with the technology initiative at the risk of being fired; they can implement the technology at the risk of losing their identity; or they can comply with the new technology, while simultaneously undermining it, thereby avoiding becoming smaller versions of themselves at work.

“[The latter] is what you see happening most of the time,” Narayandas says. “People don’t want to get adversarial or be identified as the curmudgeon who refuses to do things, so they adopt the technology symbolically, giving the impression they are using it.”

This symbolic adoption is an identity-preserving response: Users are trying to avoid a version of work in which their own expertise, role meaning, and professional value feel diminished.

As passive-aggressive as that may seem, it reflects a clear-eyed read of what employees personally stand to lose, says Zhang.

"It's not just that 'my job will be replaced and I'll receive less compensation,'" she says. "There's also an intangible return. People are asked to put in the effort of learning to work with AI, and what they get back is a smaller sense of who they are in the job. For them, that trade just doesn't add up."

Reassuring anxious employees

Despite the challenges, Narayandas and Zhang say technology projects don’t need to be doomed from the start. When approached with sensitivity, those selling the tools—and the companies they work with—can take steps to make the changes less threatening and more beneficial for approvers and users of the technology.

The point is not simply reassurance; it is to redesign the value proposition so users can see how adoption preserves or enhances the identity, expertise, influence, and future value of their role.

The researchers proposed that sellers design what they call “identity-compatible advantages,” a bundle of five complementary mechanisms that help approvers and users adopt innovative technology without feeling their roles are being undermined:

Recharter roles

Even if AI takes over higher-order analysis and reasoning, companies can show employees a more valued version of their roles. The promise cannot be simply that they will “do more strategic work.” “You have to specify which forms of judgment, customer knowledge, interpretation, problem framing, or relationship expertise become more visible because routine work has been automated,” Narayandas says.

Build decision guardrails

While AI takes over some decision-making functions, companies can make sure managers and their employees retain some discretionary oversight by overriding or questioning choices. That way, if AI makes a mistake, the employee can take over. “It’s a proverbial red button you can punch when things go out of whack to make the treadmill grind to a halt,” Narayandas says.

Create analytical overlays

Rather than replacing individual judgment, AI can provide input to enhance employees’ decision-making processes. The overlay should make the user’s expertise more visible, not simply make the AI’s recommendation easier to accept. “It’s like having a research assistant who can do the background work, but you are still the one bringing meaning to it, so you can feel good about that,” Narayandas says.

Develop redeployment pathways

Companies can make credible commitments to workers that they will be retained—and more importantly, retrained—in new roles. “The pathway matters because users need to see not only that they still have employment, but that the post-adoption role is one they can recognize as meaningful, respected, and worth growing into a better, brighter future,” Narayandas says.

Fold in executive sponsorship

Rather than leaving employees to defend an unfamiliar way of working on their own, senior leadership should publicly stand behind the redesigned role and the people stepping into it. “When leaders put their name on the new role, they're signaling it's a real, respected job—not a quiet demotion,” Narayandas says.

The approaches must be used in tandem to ensure that those responsible for implementing the technology enthusiastically buy into a new project, the researchers say. Role rechartering, for example, can’t happen without redeployment pathways that prepare someone for new tasks. “You can’t just turn salespeople into consultants,” says Zhang. “You need the training to provide those pathways, and executive sponsorship to make that happen.”

Ultimately, the researchers say, companies need to think about more than the ROI of technology and consider that employees may be numbed by fear about losing their identity or their very job.

“The machine doesn’t need to be impressed or convinced or motivated, but the human needs to understand how they can benefit,” says Narayandas. “The more managers spend time thinking about that, the more likely the technology will be adopted.”

Illustration by Ariana Cohen-Halberstam with images from Adobe Stock.