Top executives are buying into artificial intelligence in a big way, and line employees are enthused about AI. That’s all good. However, the managers in the middle appear to be, well, caught in the middle of the AI whirlwind. And this can derail AI efforts before they even get out the gate.

That’s the takeaway from a survey of 2,603 employees in organizations across the globe, just released by Infosys. Only 22% of middle managers in the survey are actively involved with AI-based transformation, compared to 49% of senior executives and 25% of junior-level employees. One in five middle managers are even reticent about sharing details about their AI usage with their organization, either downplaying or overplaying their use of it.

“Middle managers are most likely to downplay their AI usage – more than twice as likely as senior leaders," the report’s co-authors, Saman Masood and Priyanka Haldipur, both with Infosys, point out. "More than half of those downplaying their AI use say they do so because they are seen as less skilled or capable.”

A well-functioning AI effort is not possible without engagement from all levels, observers point out.

Dan Shapiro, CEO of Glowforge and Wharton Research Fellow, calls for a “200% option ” which takes organizations above and beyond the norm for AI these days. This means developing all managers and employees as “AI superheroes,” capable of accomplishing twice the level of tasks and innovations.

While the common formula is to cut employment as AI is brought in, emphasizing managers’ and employees’ AI opportunities “means doubling your output instead of halving your costs,” Shapiro says. “Expand your product line, trounce your competitors, improve quality, increase customer satisfaction, and grab market share.”

Middle managers need to be part of this 200%. But they are being left out, the Infosystems survey finds. Almost a fifth, 19%, worry they will be held responsible if AI goes wrong, which is twice the proportion of senior executives who feel that way. Only 13% of middle managers say they have the freedom to choose how they use AI, compared to 22% of senior executives. Plus, only 13% of middle managers say they are always encouraged to experiment with AI, compared to 21% of senior leaders.

In addition, only nine percent of middle managers actually reviewee incentives for their AI usage. Only 25% receive peer recognition for their AI accomplishments, compared to 40% of executives and junior employees.

“Middle managers can help and organization be more agile – apply the lessons from its mistakes and shifting its operating model quickly to grab new opportunities," Infosys’ Masood and Haldipur state.

Technology may serve as a great equalizer, “but what it doesn’t equalize is your people – their judgment, their taste, their ability to know a good answer from a plausible one, the trust they build in a single conversation,” Stephen Klein, founder and CEO of Curiouser AI recently posted on LinkedIn. “The enterprise that believes in its people spends the AI dividend on them: keeps them in the loop, hands them the tools as instruments, protects the practice that produces judgment. For the first time, the high-road strategy and the high-return strategy are the same strategy.”

The Infosys survey points to a major gap in this thinking, short-changing the managers who make the close-to-the-ground decisions that will make AI work.