Eapen et al published a short practitioner piece in Harvard Business Review in July 2023, right at the beginning of the generative AI wave, arguing that AI’s biggest opportunity is as a creativity amplifier. The paper offers a conceptual framework for thinking about how AI can support creative work, and a set of assumptions that are worth revisiting now that we’re three years into the experiment they were imagining.
The question driving the research is: can AI make people more creative, or does it just make the output look creative? Eapen et al. land firmly on the optimistic side. They argue that generative AI’s “greatest potential is not replacing humans; it is to assist humans in their individual and collective efforts to create hitherto unimaginable solutions” (p. 2). That’s a strong claim. And in 2023, it was a necessary one. The dominant narrative at the time was panic about job replacement. This paper tried to reframe the conversation around augmentation, and I think that reframing was valuable.
Four Creativity Challenges AI Can Address
The paper’s core contribution is identifying four specific barriers to innovation that AI could help organizations overcome.
Eapen et al. call the first challenge “evaluation overload.” When companies use crowdsourcing or open innovation to generate ideas, they get flooded. Hundreds or thousands of submissions pour in, many of them incomplete, and there’s no efficient process for sorting, merging, or evaluating them. Good fragments get lost because nobody has time to find them.
The second challenge is what the authors call “the curse of expertise.” Domain experts are excellent at generating feasible ideas, things that can actually be built. But that same expertise becomes a filter. It screens out novel ideas that don’t fit existing mental models. The very knowledge that makes experts productive also makes them resistant to surprise. I find this one of the paper’s strongest observations. It applies directly to education: teachers who’ve been teaching a subject for twenty years may struggle to see how AI could change what’s possible in their classroom precisely because they know so well what’s always worked.
Eapen et al. describe the third challenge as an inability to translate. Non-experts may recognize a promising idea but can’t articulate the technical details needed to make it workable. They see the potential but can’t build the bridge from concept to design.
The fourth is the difficulty of synthesizing competing requirements. Organizations struggle to combine diverse customer needs into a single coherent solution. The parts don’t add up naturally, and the act of combining them requires a kind of creative integration that is hard to systematize.
Five Ways AI Supports Creative Work
The paper then proposes five ways generative AI can address those barriers: promoting divergent thinking by helping users associate remote concepts, challenging expertise bias by generating atypical designs that push people beyond their assumptions, assisting with idea evaluation by making vague ideas more specific and testable, supporting idea refinement by merging and combining large numbers of raw submissions into stronger proposals, and facilitating collaboration between designers and end users by serving as a bridge in co-creation processes.
These are useful categories. They give educators and organizations a vocabulary for talking about AI as a thinking partner. And the divergent thinking argument, that AI can help people make connections they wouldn’t have made on their own, tracks with what I’ve observed in classrooms where students use AI for brainstorming before committing to a direction.
Reading This in 2026: What the Paper Didn’t Anticipate
This is where temporal context becomes essential. Eapen et al. wrote this piece when ChatGPT was barely six months old. The models were less capable, institutional experience with AI was almost nonexistent, and the research base on cognitive effects was just beginning to form. Three years later, we know things they couldn’t have known.
The paper assumes that humans will remain the primary thinkers and that AI will serve as a supplement. That assumption hasn’t held up uniformly. Shaw and Nave’s 2026 work on cognitive surrender shows that many users progressively hand over reasoning to AI systems without realizing they’re doing it. The creativity isn’t augmented. It’s outsourced. Eapen et al. describe AI as a “powerful collaborator, augmenting collective human creativity” (p. 1), but collaboration requires that both parties contribute. When one party does all the generative work, it stops being collaboration and becomes dependence.
Fan et al. (2025) documented this in a writing context: students using ChatGPT produced better essays but showed no additional learning gains. The product improved. The process didn’t. If we apply that finding to Eapen et al.’s creativity framework, the implication is uncomfortable. AI might help an organization generate a better set of ideas, but the people in that organization may not become better thinkers as a result.
The evaluation overload gets solved. The curse of expertise gets bypassed. But the cognitive muscles that make expertise valuable in the first place may atrophy, which is exactly what Kosmyna et al. (2025) found at MIT when they measured brain activity during AI-assisted writing.
The paper also doesn’t address the homogenization risk. If every organization uses the same AI tools to generate and refine ideas, the outputs converge. Divergent thinking is only divergent if the inputs and processes are genuinely varied. When everyone runs their brainstorming through the same model, you get the illusion of creativity without the substance. This is a concern that Kalantzis and Cope (2025) raised in their work on literacy and AI: when tools standardize the production process, the range of what gets produced shrinks, even if each individual output looks polished.
I don’t want to be dismissive of Eapen et al.’s framework. The four challenges they identify are real, and the five AI applications they propose are conceptually sound. For educators, the most useful parts are the “curse of expertise” concept and the idea evaluation application. Teachers can use AI to surface ideas they wouldn’t have considered, to combine student submissions into stronger group proposals, and to generate provocative starting points that challenge assumptions.
But the framework needs a pedagogical layer the authors didn’t provide. Every one of their five applications can go wrong if the human stops doing the cognitive work. Divergent thinking prompted by AI is valuable when the student evaluates, selects, and builds on those ideas. The moment a student accepts AI suggestions uncritically, that value disappears. And idea refinement only works if the human is shaping the final product, not nodding along as the AI polishes everything into sameness.
The paper gave us a good conceptual map in 2023. In 2026, we need to add the terrain, the cognitive research, the classroom evidence, and the hard questions about what happens to human creativity when AI is always in the room.
The tools will keep getting better. The question is whether we will too.
References
- Eapen, T. T., Finkenstadt, D. J., Folk, J., & Venkataswamy, L. (2023). How generative AI can augment human creativity: Use it to promote divergent thinking. Harvard Business Review, 101(4), 56-64. https://hbr.org/2023/07/how-generative-ai-can-augment-human-creativity
- Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X., & Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56(2), 489–530. https://doi.org/10.1111/bjet.13544 //
- Kalantzis, M., & Cope, B. (2025). Literacy in the time of artificial intelligence. Reading Research Quarterly, 60, e591. https://doi.org/10.1002/rrq.591
- Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing tasks. MIT Media Lab. https://www.media.mit.edu/publications/your-brain-on-chatgpt/