Managers and Executives Disagree on AI—and It’s Costing Companies

Sometime since the launch of ChatGPT in late 2022, most large organizations have crossed a threshold: AI moved from something they were considering to something they were officially committed to, with large budgets and bullish predictions. The question stopped being whether AI was going to transform businesses and became when the results would arrive. For three years we’ve tracked AI adoption through the Wharton School and GBK Collective Enterprise AI Adoption Study, surveying business leaders at U.S. companies with revenues above $50 million. Some of the numbers are impressive: More than 80% of the leaders report weekly use, up significantly year-over-year; the strong majority plan to expand investment; and 74% report perceived positive returns on early deployments. These business leaders are genuinely engaged and optimistic based on their initial experiences. That said, results have not materialized at a scale commensurate with investment. On that front, some surveys are discouraging. Researchers at BCG, McKinsey, and MIT, for example have converged on roughly the same figure: Less than 10% of companies are capturing meaningful AI value at scale. Only a small minority of firms, it appears, have achieved large-scale P&L impact. Our research points to a specific tension arising within organizations that could explain this problem: Executives and middle managers aren’t on the same page. What’s holding up progress? Responses vary. Some observers point to the technology’s current limitations. Others focus on the gap in use and attitudes between AI-bullish executives and AI-anxious frontline workers. For instance, a recent study from the HR consulting firm Mercer shows that 43% of workers believe AI and other advanced technology will make their job more frustrating or difficult, and 53% agree that it will impact their job security. One of us (Stefano) recently captured some of this dynamic in “Why Gen AI Feels So Threatening to Workers” (Harvard Business Review, March-April 2026). But our data reveals a more consequential divide within the leadership ranks themselves. On one side are senior executives, and on the other are middle managers, who are responsible for leading the workforce through this transformation by translating strategy and investments into actual results. The space between them is the “messy middle.” When we dissect the averages and look at these groups of leaders separately, we find they aren’t operating in the same reality. They hold fundamentally different views on whether AI is delivering ROI, how quickly their organizations are adopting AI, and how the technology may impact their teams and skills. Why is this important? Because AI adoption can be slower if senior leadership assumes that the rest of the organization shares its optimism, or if the organization underestimates the difference in expectations between the leaders setting the strategy and the middle managers responsible for its execution. Until executive leaders understand this divide and take ownership of closing it, no level of investment or ambition at the top is likely to produce the transformation they’re promising. The Evidence Our study found significant gaps on the most basic question a business leader should be able to answer: Is your AI investment delivering returns? Senior executives and middle managers gave us starkly different answers to that question. Nearly half of executives (45%) reported significantly positive ROI from their initial AI investments. Among middle managers, that number drops to 27%. That 18-point gap in the fundamental assessment of whether the strategy is delivering results illustrates a potentially costly disconnect. The divide even extends to how the two leadership groups perceive the pace of change itself. We asked whether their organization was adopting AI faster than competitors, 56% of executives said yes, “much quicker.” Among middle managers, only 28% agreed. The sentiment data are equally telling. Over the past year, nearly two-thirds of executives say they have become “much more positive” about generative AI, a remarkable shift in a single year. Among middle managers, only 39% say the same. Middle managers are 64% more likely than their senior colleagues to describe themselves as “cautious” (46% vs 28%). That said, they’re not opposed to AI; more than half say they are “excited” and 62% are “optimistic.” These findings are consistent with other research. A BCG and Columbia Business School survey of nearly 1,400 employees and leaders found that 76% of executive leaders believe their employees feel enthusiastic about AI adoption, whereas only 31% of individual contributors say they are. And where in the mix are middle managers? Almost exactly in the middle, at 51%—25 points below where leaders are. Why do executives and middle managers have different takes? Part of this divergence reflects something real about the nature of executive versus managerial work. Senior leaders tend to use AI for high-level synthesis, strategic drafting, and decision support, tasks where the technology performs well, so the current capabilities tend to benefit their work. Middle managers, on the other hand, are tasked to deploy the technology in the messier territory of day-to-day operations: workflows built over years, teams with uneven technical comfort, output that has to be consistently right, not just fast. When the tool works, both groups understand and reap the benefits. When it fails, typically only one of them has to cope with the aftermath. And for those middle managers who already feel overworked and under-resourced, learning to use yet another tool can feel like an unwelcome additional burden rather than a helpful copilot. These two groups are also operating on fundamentally different time horizons. Executives are rewarded for vision, which is to say for seeing around corners and betting on what’s possible; middle managers are rewarded for execution, which is to say making things work today, with whatever people and processes are at hand. That difference matters. Middle managers are close enough to AI’s current limitations—the hallucinations, the integration friction, the day-to-day workflow disruption—to hold a more grounded view of what it can and cannot do right now. Executives are free to focus on where it might be going. Bridging that gap is the key to transformation. Missing the Middle It would be easy to read these gaps as simply the difference between two types of leaders: visionary executives focused on strategic possibility versus pragmatic middle managers focused on operational reality. There’s something to that framing, but it’s insufficient. It suggests the gap is natural, maybe even healthy: two necessary perspectives in productive tension. We disagree. We believe that this tension, if left unaddressed, could unnecessarily delay organizations’ ability to transition from AI ambition to AI value. Whatever executives aspire to, it is middle managers who will actually make it real—or not. They live in the messy middle where the work gets done, where most of the workforce gets led, and where AI has the most potential impact on jobs, perhaps even their own. If middle managers proceed more cautiously or reluctantly than senior leaders want them or expect them to, frustration and tensions are bound to emerge on both sides. The divide highlighted in this article may also suggest that senior leaders struggle to understand how AI affects workflows and day-to-day work, potentially undermining the quality of decision making. “Leadership’s job is to help move from ‘AI is possible’ to ‘AI is practiced’,” we were told by Ana White, the chief people and AI enablement officer at Lumen. “Most successful transformations aren’t top down—they’re won or lost in the day-to-day realities that managers navigate under real constraints. AI transformation has to reflect cultural and human realities, not just executive ambitions.” There is a compounding challenge underneath this. McKinsey research has shown that managers typically spend less than 30% of their time on the talent and people leadership tasks that are now becoming one of their most critical functions in this transformation. Nearly half of their time goes to administrative work and their own individual contributor responsibilities. Gen AI holds genuine promise to return some of that time—for example, by automating status updates, summarizing meeting notes, and drafting routine communications. But capturing that promise requires that managers learn to use new tools, adopt new workflows, and reorient their teams. Leaders that hand middle managers an AI mandate without first reducing their administrative load and providing them with necessary resources are asking them to build the plane while flying it. “The executives I see succeed at this,” Christine Grimm, a longtime executive coach and founder of Aria Consulting, told us, “do something uncomfortable: They stop asking why their managers aren’t moving faster and start asking what they themselves haven’t done to make it possible. That’s a hard question. It’s also the right one.” Five Concrete Actions to Take Our evidence shows a potentially costly fault line between executive leaders and middle managers. Getting those groups on the same page about their organizations’ AI transformation is not simply a communication exercise. It requires executives to acknowledge that the conditions they have created—or failed to create—will determine whether their AI aspirations make it off the ground. The solution to this problem isn’t more technology investment or bolder vision. The organizations that break through will be the ones whose executive leaders turn their attention inward to the managers carrying the burden of this transformation and equip them with sufficient support, clarity, and bandwidth. Cleaning up the messy middle and closing this fault line starts with five concrete actions leadership teams can take now: 1. Diagnose before you prescribe. Where is your organization on its AI transformation journey? Do managers and teams understand and embrace your AI vision? Are there pockets within your organization that are more ready or more resistant to these changes? You can’t close a gap you don’t see, and without intentional evaluation, executive leaders’ enthusiasm may muddy their grasp of their organizations’ realities. 2. Co-create the playbook, don’t hand it down. Optimistic executives should promote their vision and aspirations but need to listen to their managers’ practical concerns and input. Bring managers into the roadmap discussions before decisions are made, not after. The goal is to ensure all levels of leadership are aligned and enlisted on a common path—not just aware or engaged. 3. Sequence the effort by reducing load before adding to it. In today’s demanding and dynamic workplace, managers are increasingly trying to keep the trains running on time. Successful AI transformation requires re-thinking workflows, re-skilling team members, running intentional test and learn pilots, and learning how to manage hybrid AI agent and human teams. These incremental time investments are required before the organization can achieve the desired future efficiencies and growth. Executive leaders need to recognize these necessary near-term burdens and work with their management teams to ensure sufficient capacity to accommodate them. 4. Measure readiness, not just adoption. While adoption remains the goal, tracking usage rates alone will miss the mark. Leaders need to ensure they have a deeper understanding of their teams’ readiness—that is, not only their relevant skills and capabilities but also their attitudes and perceptions. Manager confidence and organizational readiness should be explicit KPIs alongside usage metrics. 5. Build upward feedback loops and reward honesty. Many of the gaps we’re seeing indicate a critical need for middle managers to align with their executive leaders. Devise shared KPIs to guide discussions on what’s working and what’s not. Create structured channels for middle managers to report implementation reality upward and make clear that cautious assessments or failed pilots are valued data, not resistance. . . . When will meaningful results finally arrive? Only time will tell, but we believe the organizations that close this fault line and clean up the messy middle will be the first to find out.

Managers and Executives Disagree on AI—and It’s Costing Companies

Summary.

Partner Center