Despite significant investments and widespread optimism, most organizations find themselves struggling to unlock the transformative potential of artificial intelligence. As one illustration of the challenge, a BCG study in late 2025 found that just 5% of companies achieve substantial value from AI, while 60% report no material value at all. Leaders are puzzled and frustrated, as their expectations remain high. So why are so many AI initiatives falling short?
AI-driven change is fundamentally human in nature. When AI-driven change efforts fail, they usually do so due to human factors including poorly understood goals, insufficient buy-in from employee segments, underwhelming take up, and uninspiring storytelling, to name just a few. When AI-driven change succeeds, it is because leaders put people at the heart of every decision, informed by a rich understanding of their emotions, thoughts, and behaviors. BCG has observed that, in successful AI-driven transformations, 70% of the value is derived from people-related action rather than technology-related action.
Seven essential principles from behavioral science explain the difference between efforts that succeed and those that do not. They have nothing to do with what a machine can do. They are entirely about what the people—leaders and employees—think, feel, and do. We have spent years distilling these principles into a book that tells the stories of companies that have managed to successfully navigate change. We wrote about them extensively in our forthcoming book How Change Really Works, which identifies the seven principles for successful change, informed by half a century of behavioral science findings from around the world.
Stay ahead with BCG insights on people strategy
Principle 1: Get True Agreement, Not False Alignment
Change programs often get off to a false start. This happens because a company’s leaders haven’t truly agreed on the purpose of the change. This is particularly the case with AI transformation because different leaders have different hopes and expectations for what the new technology will deliver—everything from cost reductions and improved productivity to new innovations and higher growth. They may all seem to agree with each other, but they have not properly resolved the underlying trade-offs—they have settled for what we call “false alignment.”
Executives need to truly agree on where AI interventions actually matter—not pursuing “AI everywhere,” but instead focusing deeply on a few critical workflows where AI fundamentally changes the economics. BCG research demonstrates that successful companies prioritize just three or four use cases on average, compared with six or seven for less-successful organizations.
To achieve this, executives reach shared clarity on their underlying intent. Some may aim for immediate scale and tangible value creation. Others might focus on long-term innovation, with success measured through new capabilities or strategic positioning. Others still could simply be investments in building AI readiness—helping employees gain familiarity, maintain relevance, and build comfort with the technology—so they’re prepared to tackle more complex, high-value applications when the time comes.
True agreement also means achieving clarity on how success will be measured, the timeframe to meet those targets, which leaders will take responsibility for which parts of the program, and how further decisions will be made as the program plays out, including decision-making around pilots.
Finally, a true agreement on AI-driven change must include clear principles for investment that match the program’s stated ambition. This is where the rubber hits the road, especially where leaders have agreed (and perhaps announced) that AI is their company’s top priority. Tough reallocations of funding and resources, including top talent, are commonly required.
Principle 2: Increase Agency, Not Just Involvement
Managers need meaningful agency in designing new AI-enabled workflows for their teams, precisely because their own roles are often the most significantly disrupted. In many traditional workflows, middle managers primarily synthesize across teams, coordinate handoffs, and manage throughput. But AI fundamentally changes that equation: it streamlines coordination and reduces administrative workloads, shifting managers’ focus away from managing process toward making higher-value decisions about substance—real design choices for their products and services.
In an expression of the IKEA effect, managers who have some agency in defining how their own roles will evolve are more likely to perform those roles effectively from day one. This means providing opportunities for managers to make meaningful decisions in the design of new AI-enabled workflows and have tangible influence over decisions that others make. When managers are offered a clear stake in shaping their new roles, they’re more prepared, motivated, and capable of successfully guiding their teams through a transition.
But beware: managers cannot effectively exercise their agency without strong programmatic support. Managers benefit greatly from both dedicated time to explore and experiment with potential uses of new technology for their teams and structured opportunities to learn from other, more experienced managers who have had tangible success.
Principle 3: Expect Take Up to Be Earned, Not Automatic
Earning take up is a challenge that every change—not just AI-driven change—must face. There are seven barriers that can prevent employees from successfully taking up new behaviors. All of them are relevant for AI adoption. Three barriers are especially common:
First, a skills gap frequently arises, and it is broader and deeper than many leaders might assume. It spans far more than the habit of delegating tasks to AI or rigorously checking the outcomes; it encompasses the full range of capabilities employees need in order to work effectively with AI in their specific roles. BCG’s recent AI at Work report found that only 36% of employees say they have been trained on the skills needed for AI transformation. This means that nearly two-thirds are likely underprepared for their adoption efforts. Closing this gap also requires rethinking how skills are developed, not just how much time is dedicated to the effort. Leaders can certainly help by providing protected time for employees to practice and experiment. But the approach to active skill-building must change as well, moving beyond tactical training toward learning that is sustained, role specific, and deeply connected to each employee’s actual work.
Second, employees often perceive a permission gap for AI use. They may feel uncertain whether using AI is culturally acceptable, or if colleagues might label it as “cheating.” In addition, they may have cybersecurity and copyright concerns. Leaders can address this by consistently role modeling AI usage in their own work to prevent any gap between behavior they are requesting from others and behavior they are personally demonstrating. Every leader should be able to describe one way they applied AI in the previous week. They should know their personal answer to the question, “What’s your favorite AI use case?” In addition to role modeling, clear expectations on AI use should be introduced in performance management reviews.
Third, employees can feel like they have something to lose—particularly their professional identity and pride in their craft. People often worry that AI could diminish the expertise and tasks they genuinely enjoy, such as writing code, designing products and services, building an advertising campaign, or developing innovative new strategies. Leaders can help solve this by speaking often, and passionately, about how AI elevates, rather than replaces, the deeper purpose behind their work—enabling better code, sharper designs, more effective campaigns, and smarter strategies. It also helps to give these employees opportunities to co-create the quality bar for AI-enabled work product, rather than having it imposed. But understand that this barrier can be the hardest to reduce: for some employees, a mindset change of this magnitude may take many months.
Principle 4: Understand Emotions Through Feedback, Not Instinct
Employees’ emotions about AI have a significant impact on their behavior around AI, which drives the effectiveness of AI-related change programs. Many employees feel fear—that their colleagues will disapprove of their extensive AI use, or that they are not as fluent in AI as they are expected to be, or that AI-supported work they share may be perceived as low quality, or that their roles will ultimately be replaced by AI. In all cases, fear restricts candid dialogue and delays meaningful progress. Other employees feel grief over losing their professional identity, expressing resistance through exaggerated “quality concerns” that result in unnecessary skepticism or rejection of useful AI outputs. Others still feel sadness, as they perceive that human beings are losing the agency to make their own decisions in their own organizations.
These negative emotions stand in sharp contrast to the enthusiasm executives often feel about AI adoption: a recent BCG survey found that 76% of executives believed that employees were excited about AI, yet only 31% of employees actually felt that way. But even when employees do share executives’ positive feelings, excitement itself can create issues. Excitement without clear structure often leads to superficial, “shiny-object” experimentation, causing employees to briefly try many different AI tools without mastering any and then quickly revert to old habits—undermining lasting change.
To accurately gauge employees’ emotions around AI, leaders can’t rely solely on their instincts—they must measure consistently. Frequent, short “pulse surveys” can track employees’ emotions, confidence, and capacity in real time. If formal surveys are likely to face cultural resistance, alternatives such as AI-led interviews can help leaders gather accurate measurement on the prevailing emotions across their workforce, while fostering familiarity with AI tools through the experience. Employees’ emotions around AI are shifting rapidly as the technology evolves, so pulse surveys must be frequent and leaders must be willing to revisit their assumptions in light of the latest evidence.
Principle 5: Use a Process with Rituals, Not Reactions
During AI-driven transformation, leaders must carve out structured, predictable time—ideally every one to two weeks—to get into the details of execution. Regularly scheduled, ritualized sessions ensure that AI-driven changes remain sharply focused on what truly matters, even as the organization experiments and learns what is possible. Such a disciplined approach replaces ad hoc reactions with intentional reflection and decision-making, helping leaders avoid decision fatigue and maintain clarity and pace amid complexity.
In successful AI transformations, leaders follow rituals that nudge them to debate the hardest questions. For example, what did we learn from the latest changes we made to critical workflows? In light of those lessons, do we still have the right teams and structures in place? Where should we raise our formal targets for initiatives that are delivering better than we expected? Which initiatives should we stop altogether? Where do we urgently need to find more capacity and resources? And what will we prioritize next?
Leaders also use their shared rituals to build a culture of commitment to company-wide AI proficiency. They develop structured traditions of sharing their personal AI successes—and challenges—as individual users of the technology.
Principle 6: Share Stories and Symbols, Not Just Dollars
There are three types of stories that leaders can use to effectively explain to employees the motivation for change. A threat story highlights what’s at risk if the organization doesn’t change—highlighting potential losses or negative consequences. A fitness story focuses internally, stressing the need to improve, become more efficient, or strengthen capabilities. A destiny story draws focus to the special qualities that a company possesses and argues that, through change, the market can gain more or better access to those special qualities.
Some leaders default to a fitness story when deploying AI, communicating to employees that the goal is to do the same work they always did—but more efficiently. The unintentional implication can be that people aren’t working hard enough or efficiently enough, making the change feel like a criticism rather than an opportunity. Employees often experience this as added pressure rather than motivation to embrace new ways of working.
A more effective approach is the destiny story, one that says, “This is about scaling your expertise and passion far beyond what was previously possible. AI doesn’t just save you a few hours—it empowers a single person to accomplish tasks that once required an entire team.” This hyper-leverage narrative generates curiosity, not fear, by focusing on what employees can uniquely achieve, not simply what they can do faster.
Principle 7: Create Momentum Throughout, Not Just at the Start
Individuals become more motivated and engaged to reach a goal when they can see that some tangible progress has already been made. Scientists call this the endowed progress effect. Constantly communicating and celebrating small wins reinforces this feeling of ongoing progress, creating sustained momentum and helping people maintain their motivation over the long term. Early momentum does not continue indefinitely; in fact, MIT’s Project NANDA recently reported that only 5% of custom enterprise AI tools reach production.
For AI-driven changes, there is an abundance of small wins to find and promote. Leaders can highlight a team that eliminated an entire step from a cumbersome workflow—not just saving time but also completely removing a frustrating task. They can recognize and publicly celebrate an individual who, using AI, accomplished something previously thought impossible, such as one person delivering what used to require a team of five. They can gamify individuals’ AI achievements and developing competencies. They can emphasize concrete business metrics that moved as a result of AI, such as a reduction in cycle time without compromising quality—not vanity metrics like AI usage statistics. These tangible, relatable examples sustain energy and show teams that progress is real and meaningful.
To be most effective in promoting wins, it is important for leaders to highlight individuals across the proficiency spectrum—not just employees in the top 1% of AI superusers. Highlighting relatable peers is often effective in motivating employees to engage and progress further.
In an era captivated by a profound new technology, we must not lose sight of what has driven every technological leap forward: people. To realize AI’s full potential, leaders must place human beings at the center of every change—understanding their emotions, addressing their behaviors, and carefully shaping their experiences. By applying the seven science-based principles outlined in this piece, companies can give their AI-driven changes the best possible chances of lasting success.
The authors thank Gabriella Kellerman for her help finalizing this article.
