Managers aren’t ready to lead the AI talent they’re hiring

“The team members are doing all those discrete tasks, and then the managers need to think about how to organize this kind of AI-enabled workflow,” Wen said. “This is probably one of the reasons managers are behind their teams, because they’re managing a much more complex and uncertain problem.”

The evaluation problem

Wen’s research, including a recent Harvard Business Review study she co-authored on AI-enabled workflows, found that when employees rely heavily on AI, managers lose visibility into what those employees are actually contributing.

“What you see is not really what the employees do individually. It’s really about the output from this human-AI collaboration,” Wen said, making it harder to evaluate performance or decide who deserves a promotion.

In the past, Wen said, managers could look at a report or a deliverable and reasonably assume it reflected an employee’s own thinking. Now, an employee might hand in polished work that’s almost entirely AI-generated, and it can look just as strong as work someone did largely on their own. To separate the two, Wen said, a manager would first need a clear sense of what AI alone can produce, then work backward to figure out what the person actually added.

“Those tasks are very hard to even assess,” Wen said, because most organizations don’t have a clean benchmark for what AI can do on its own in a given role. Without one, she said, managers are often left guessing at how much of an employee’s output reflects real skill, which complicates decisions about promotions, team structure and how work gets assigned in the first place.