Generative AI Has a Big Scary Side Effect. Here’s the Cure.
By Chris Weller and David Rock Imagine you’re the CEO of a company. You just hired 100 new managers whose only job is to walk around the office and pop into employees’ chat windows to give them answers to every leadership problem they face. Now let’s ask: Would that be a good thing? Hopefully, you said no. But that’s essentially what organizations are doing when it comes to artificial intelligence software: hiring a question-and-answer machine to give employees the impression that their job can be offloaded quickly and easily, no critical thinking required. Every leader wants their teams to have access to as much information as possible, as quickly as possible, because they believe in a lofty promise of generative AI (GenAI). That is, GenAI will allow their employees to skip the menial everyday tasks that eat up their time and energy, and focus solely on high-value work. However, research suggests that overuse of AI, in the wrong ways, can promote lazy, uncritical thinking and overconfidence in impractical solutions. So how can leaders ensure that team members are using AI most effectively? Our research into GenAI, including building our own natural-language AI leadership coach, suggests there are four domains of work that make up the four “bases” of a culture’s DNA, and which AI can easily plug into: coaching, mentoring, feedback, and role playing. When teams use AI more deliberately to support these essential functions, they stand the best chance of enhancing—rather than diminishing—the organization’s outputs overall. Accelerate Insights through Coaching Let’s say you’re stuck on a task. Ideas aren’t coming to you, so you consider asking AI for its thoughts. Science suggests this is the wrong approach for coming up with solutions that are truly novel. In fact, the smarter approach is to ask AI to prompt you. Research suggests this is how you’ll arrive at the crucial experience of insight—that “Aha” moment when a new perspective comes into focus and you feel a jolt of intrinsic motivation. Consider one recent study, which found that when people used AI to help them write stories, the stories they produced were more creative than the ones they wrote on their own, but the AI-assisted stories were also quite similar to each other. In a classic case of regressing to the mean, AI made the work more creative individually but less novel overall. To accelerate the experience of insight, allow AI to prompt your own thinking. Ask it to ask you how else you might look at this problem, what kinds of solutions you haven’t tried yet, and how you might rearrange existing ideas to create innovative combinations. Science shows that learning happens in the “zone of proximal development,” or ZPD. This is the sweet spot between what you can do without anyone’s help and what you can’t do even with someone’s help. When AI helps us expand our ZPD, rather than spoon-feeding us generic suggestions, we become smarter and more creative because we can see new connections we previously missed. Get Expert Advice From a Mentor Leadership theories and styles number into the hundreds, maybe thousands, which means there are bound to be confusing contradictions. How is a leader supposed to know which theories to follow? At the NeuroLeadership Institute, our recommendation is to follow the neuroscience, which presents one coherent theory for behavior based on how the brain works in various conditions. AI plays a major role in following science’s lead. Instead of your team members asking AI for generic solutions to challenges they face, encourage them to ask more pointed questions framed around what the research says about the specific issue. For instance, instead of asking “How can I deliver negative feedback better?” you’re more likely to get a proven solution, backed by research, by asking “What does the research say on how to best deliver negative feedback so the other person will listen?” By anchoring on science, you’ll get a more complete and coherent picture of how different principles fit together. You’re getting smarter in unfamiliar situations because the principles stay the same: Instead of following 100 thought leaders, you become more fluent in one consistent body of knowledge. When AI offers empirical suggestions, it doesn’t just solve a person’s current problem; it creates a foundation of knowledge, rooted in first principles, that can be applied in future situations. Used properly, AI can help us make the important leap from knowledge to wisdom. Receive Real-Time Feedback In the spirit of co-opting AI to become a thinking partner, not just an ATM for solutions, AI can “sit in” on meetings and presentations and listen for specific patterns you may want to fix or improve. This allows you to engage in a process known as metacognition: focusing not just on the contents of your problem, but asking AI to parse your speech and detect clues that might signal something about your patterns of thought. In that search, AI may reveal unhelpful assumptions or unconscious biases that are affecting your thinking or decision-making. For example, an employee may ask AI to help them navigate an interpersonal conflict at work. Along with giving the backstory and the problem, the team member may consider including one additional prompt: “Is there anything about the way I’ve framed this problem that might be limiting how I think about it?” Such a level of self-awareness expands the parameters for AI’s feedback—rather than commenting on the situation directly, it can also speak to the person’s habits of mind that might unknowingly be getting in the way. Role Play to Build Skills and Habits At NLI, we define culture as “shared everyday habits.” That is, the culture of an organization reflects how people behave—individually and with one another—on a regular basis. When employees work with an AI companion, they have an opportunity to practice and rehearse key workplace scenarios to maximize the real interaction later on. For instance, a leader who has an upcoming meeting to discuss an employee’s low performance can use AI to build the
