TLDR
- AI does not automatically boost creativity. Employees only produce more creative outcomes when they actively reflect on and refine their thinking – a skill called metacognition.
- Creativity creates a competitive advantage for enterprises in the long run, according to Shuhua Sun, a management and entrepreneurship professor at Tulane University.
- Anyone can develop metacognition. Companies could train employees to practice this skill.
Most companies rolling out generative AI tools expect a straightforward payoff: faster work and more creative output from their employees. But new research suggests that this creativity – which leads to a competitive edge over the long run – is not guaranteed. The difference? It depends on who practices metacognition.
“Metacognition simply means cognition about the cognition, which means that we reflect on our thinking process,” said Shuhua Sun, an associate professor of management and entrepreneurship at Tulane University, in an interview with The AI Innovator. “Thinking about thinking is metacognition.”
Sun and his fellow researchers discovered the importance of metacognition to creativity when they conducted a field experiment with 250 employees at a technology consulting firm. Employees were randomly assigned to either receive access to a large language model (LLM) or not. Researchers measured creativity using both supervisor evaluations and independent external raters.
The results showed that employees with access to AI produced more creative outcomes — but mainly if they used metacognitive strategies, or the ability to monitor and adjust one’s thinking. Employees who actively evaluated their approach, identified knowledge gaps and refined their use of AI were significantly more likely to benefit, according to their research paper.
“You monitor your problem-solving or thinking, your reasoning process, you see whether you are on the right track or not, and then you make adjustments,” Sun explained. “That’s called metacognition.”
By contrast, employees with low metacognitive awareness saw little improvement, even with access to the same tools. The study concludes that AI enhances creativity by providing cognitive job resources, but only for workers who use it reflectively rather than passively.
Why metacognition is important
These findings are crucial to organizations wishing to benefit from the innovative nature of AI. “The most important focus is really to boost employee creativity, because creativity leads to long-term innovation, which gives organizations competitive advantages,” Sun said.
The concept is simple but has far-reaching implications for how organizations approach AI adoption.
Since the debut of ChatGPT in late 2022, companies have rushed to integrate generative AI into daily workflows. The expectation has been that tools capable of producing text, ideas and analyses on demand would naturally lead to more creative work. But early evidence has been mixed. A Gallup survey cited in the study found that only 26% of employees reported improved creativity from using AI.
Sun’s research helps explain that gap.
The study shows that AI systems enhance creativity by providing what researchers call “cognitive job resources” — including access to vast amounts of information, the ability to process knowledge quickly and the flexibility to shift between complex and simple tasks. These resources are essential for creativity, which depends on combining ideas in new and useful ways.
But access alone is not enough. Employees must actively use those resources in a thoughtful, iterative way. That is where metacognition comes in.
Workers with strong metacognitive strategies continuously evaluate what they know, what they do not know and whether their approach is working. They question assumptions, refine their thinking and adjust their strategies as they go. In the context of AI, that often means going back and forth with the tool — testing prompts, challenging outputs and exploring alternatives.
Employees who lack those habits tend to use AI differently. They may generate an answer and accept it without further scrutiny, treating the tool as a shortcut rather than a collaborator. As a result, they are less likely to produce creative outcomes.
“Only some of the employees can generate more creative ideas when they were given the generative AI tools,” Sun said.
To be or not to be creative
The findings suggest that AI amplifies existing cognitive behaviors rather than transforming them. Workers who are already reflective and adaptive become more so with AI. Those who are not may see little change.
The distinction also highlights a gap in how employees are typically trained. Most education systems emphasize analytical thinking — breaking down problems and applying known methods to solve them. Less attention is given to questioning whether the problem is framed correctly or whether the chosen approach is appropriate.
“People who have a college education are very good at cognition or problem-solving or analytical thinking,” Sun said. “What they need to have is metacognition.”
That difference matters because solving the wrong problem efficiently does not lead to better outcomes. In fact, it can reinforce inefficiencies. “Sometimes the problem is wrong,” Sun said. “And that’s why you can’t solve the problem.”
Generative AI may exacerbate that risk. By producing quick, polished answers, it can create the illusion of understanding and discourage deeper reflection. Without metacognitive awareness, employees may move forward with flawed assumptions or incomplete analysis.
The research suggests that metacognition acts as a counterbalance. It encourages users to pause, reassess and refine their thinking, leading to more robust and creative solutions.
For example, a consultant working on a client problem might initially use AI to generate a set of recommendations. A metacognitive approach would then involve questioning those recommendations: Is this the right problem? What assumptions am I making? What information is missing? How can I improve this idea?
The employee might then use AI again to explore alternative perspectives, gather additional data or test different approaches. Through this iterative process, the final solution becomes more refined and more creative, according to Sun.
By contrast, an employee without metacognitive habits might accept the first output and move on, missing opportunities to improve the work.
AI plus humans
The implications for organizations are significant. Many companies have focused their AI strategies on selecting tools, building infrastructure and identifying use cases. But Sun’s findings suggest that human factors — specifically how employees think — may be just as important.
“Even the most advanced LLMs may fail to boost creativity if employees lack the metacognitive strategies needed to use them effectively,” the paper concludes.
That insight points to a shift in priorities. Rather than assuming that technology alone will drive innovation, companies may need to invest in developing employees’ metacognitive skills.
Sun argues that metacognition is not an innate trait limited to a few individuals. “It’s not a rare skill,” he said. “Everybody has this capability. The question is whether you are going to use it.”
That means it can be taught and reinforced. Training programs can encourage employees to ask key questions throughout their work: Am I solving the right problem? Am I using the right strategy? What assumptions am I making? What do I need to learn next? These prompts can help build the habit of reflecting on one’s thinking.
Managers and workplace culture also play a role. Environments that prioritize speed and output over reflection may discourage metacognitive practices, even if employees are capable of them. By contrast, organizations that value questioning, iteration and learning may see greater benefits from AI adoption.
Metacognitive training needed
The research also raises broader questions about the future of work.
As AI tools become more capable, the skills that differentiate workers may shift. Technical expertise will remain important, but the ability to guide and refine thinking — both one’s own and that of machines — may become a key advantage.
In that sense, metacognition could be seen as a bridge between human judgment and machine capability. AI can generate ideas, but it does not determine which ideas are worth pursuing. It can provide information, but it does not decide how that information should be used. Those decisions still depend on human thinking — and on the ability to examine and improve that thinking.
Sun’s research suggests that this ability is not just a complement to AI, but a prerequisite for using it effectively. As companies continue to invest in generative AI, the lesson may be that the most important upgrade is not the technology itself, but how people engage with it.
“They should deploy AI, but what’s also important to remember is AI will not make their employees automatically more creative,” Sun said. “They probably need to train them or enhance employees’ use of metacognitive strategy so that they can truly leverage artificial intelligence to reap value.”
