Since the advent of large language models like ChatGPT in late 2022, questions around their potential in creative problem-solving have grown louder. While generative AI has shown strength in summarizing information, identifying patterns, and generating content in existing styles, concerns remain about its ability to generate genuinely novel, innovative ideas. A new research study led by Harvard Business School Assistant Professor Jacqueline Ng Lane sheds light on this debate, suggesting that the future of innovation may not lie with AI or humans alone—but rather with both working together.
Published in the journal Organization Science, the study set out to examine whether generative AI can truly tackle open-ended, unsolved problems—particularly those that benefit from diverse perspectives and cross-disciplinary thinking. Lane, along with her co-authors from the University of Washington, ContinuumLab.ai, and Harvard Business School, explored how AI performs in comparison to human crowdsourcing in a creative context: generating business ideas for the sustainable circular economy.
The research design involved soliciting ideas from both human participants and ChatGPT. The human cohort, comprising 125 individuals from diverse professional backgrounds, submitted ideas in response to a challenge posted on an online platform. Incentivized with a $1,000 reward for the best idea, participants proposed a variety of creative concepts—such as dynamic pricing algorithms to reduce food waste and apps that digitize receipts to cut down on paper use.
Meanwhile, the researchers crafted a series of carefully engineered prompts for ChatGPT. These prompts were designed to generate a high volume of unique, relevant ideas. Some prompts were tailored to mimic specific professional personas, while others focused on satisfying criteria like environmental benefit, feasibility, and profit potential.
To fairly evaluate the quality of the ideas, the team enlisted 300 experts in the circular economy to assess submissions based on novelty, value, and feasibility. Their findings were revealing: while human-generated ideas were rated as more novel, AI-generated suggestions were more feasible and easier to implement. For example, a human-submitted idea proposed making interlocking bricks from waste plastic and foundry dust—highly original but impractical. On the other hand, an AI-generated idea to convert food waste into biogas proved to be less inventive but more likely to yield tangible benefits.
Interestingly, the most promising results emerged when humans and AI collaborated. When researchers introduced iterative prompting—asking ChatGPT to avoid repeating previous suggestions or to build upon them with different angles—the quality of ideas improved significantly. Novel concepts included everything from waste-eating flies to smart beverage containers that pay users for recycling.
Lane and her team concluded that organizations stand to gain the most from AI when they treat it as a collaborative tool. Rather than replacing human creativity, AI can serve as a catalyst—refining ideas, improving feasibility, and scaling innovation. However, the researchers cautioned against over-reliance on AI, which could stifle truly groundbreaking thinking in favor of safer, incremental improvements.
The takeaway for business leaders is clear: cultivating an AI-literate workforce and integrating AI strategically into the creative process can unlock more innovative, viable solutions. By combining the originality of the human mind with the efficiency of AI, the path forward for sustainable innovation becomes much more promising.
