Can AI be a good creative partner

What generative AI typically does best - recognise patterns and predict the next step in a sequence - can seem fundamentally at odds with the intangibility of human creativity and imagination. However, Cambridge researchers suggest that AI can be a useful creative partner, as long as there is clear guidance on how ideas should be developed together.

It's a big question for people working on creative tasks: can humans and AI work together to boost creativity? Studies have yielded widely inconsistent results in how humans working with AI can handle creative tasks together.

But a team of researchers, including from the University of Cambridge, say that while simply adding AI into a process won't improve creativity, AI can be an effective creative partner, if there are clear instructions and guidance - for both humans and AI - on how to develop ideas together.

Their research, published in the journal Information Systems Research, offers practical insight for improving human-AI collaboration to improve creativity.

"Adding AI doesn't automatically lead to better ideas," said co-author Dr Yeun Joon Kim, from Cambridge Judge Business School. "For human-AI pairs to work together and improve ideas over time, organisations must provide targeted support - such as guidance on how to build on and adapt ideas - to help employees and AI learn how to create more effectively."

In their research, Joon and his colleagues redefined 'augmented learning' - a term first used in 1962 to describe how technology can help people learn more effectively.

The researchers argue that in the age of generative AI (GenAI), learning is no longer just about improving human understanding. Instead, it's becoming a shared process where humans and AI learn and create together. The researchers describe this as an evolving partnership where both sides adjust their roles across tasks such as generating ideas, giving feedback and refining concepts.

Traditionally, technology was seen as a tool that simply made information easier to access. But GenAI, they say, acts more like a collaborator. Once a human enters a prompt, the system can take an active role in shaping ideas and decisions: shifting augmented learning from an individual process to a collective one.

The study points to Netflix as an example of human-AI teamwork. Rather than treating scriptwriting as a single task, Netflix breaks it into stages like idea generation and evaluation. Human writers create early drafts, while AI analyses character arcs, pacing and audience trends to help refine stories and improve how shows are developed and marketed.

Joon says he became interested in this research because AI was originally developed to answer a longstanding question: "Can machines generate something genuinely new?" Because while traditional technologies excel at routine tasks, many doubted that technology could make a creative contribution.

"When GenAI systems became widely available, I noticed that although they could generate ideas rapidly, people did not know how to collaborate with them in a way that improved creativity," he said.

"We wanted to figure out how people can learn to work with GenAI in a more intentional way, so that human-GenAI co-creation leads to stronger joint results rather than just more content," said co-author Dr Luna Luan from the University of Queensland.

The research included three linked studies, each involving between 160 and 200 human participants. The first study found that human-AI teams did not automatically become more creative over time when tackling social and environmental problems. The second study explored why. It identified three types of collaboration - humans proposing ideas, asking AI for ideas, and jointly refining ideas - and found that only joint refinement boosted creativity. But participants rarely increased this behaviour. A third study showed that simply instructing people to focus more on co-developing ideas led to clear improvements in human-AI creativity across repeated tasks.

"We were surprised that human-AI pairs did not naturally improve through repeated collaboration," said Joon. "Despite AI's generative power, creativity did not increase over time. We found that improvement occurred only when we introduced a deliberate intervention."

"Specifically, instructing participants to engage in idea co-development - focusing on exchanging feedback and refining existing ideas rather than endlessly generating new ideas - was the key."

The researchers say GenAI tools need to do more than churn out ideas. Their findings show that human-AI teams became less creative over time, mainly because they stopped engaging in the back-and-forth refinement that actually improves results. They say that AI systems should be designed to prompt users to give feedback, expand on suggestions and refine ideas, rather than racing through idea generation.

For organisations, the message is that creativity won't automatically improve just by adding AI. Effective collaboration requires structure: clear instructions, templates and workflows that help people recognise when to challenge, refine or build on AI-generated ideas. Training should also teach staff how to treat AI as a creative partner by practising feedback exchange and iterative improvement.

The authors also call for a shift in how companies think about human-AI work. Instead of choosing between automation and collaboration, tasks should be broken into stages where humans and AI play different roles: for example, AI generating options and humans evaluating and refining them.

They warn that although many firms rushed to adopt GenAI after the release of ChatGPT in 2022, simply using the technology does not guarantee greater creativity at work. Its impact depends on how well people understand it and how effectively they collaborate with it.

The research was co-authored by Luna Luan, a Lecturer in Management at the University of Queensland in Australia, who recently earned her PhD at Cambridge Judge Business School, Yeun Joon Kim of Cambridge Judge and Jing Zhou, Professor of Management at Rice University in Texas.

Reference:
Yingyue Luna Luan, Yeun Joon Kim, Jing Zhou. 'Augmented Learning for Joint Creativity in Human-GenAI Co-Creation.' Information Systems Research (2025). DOI: 10.1287/isre.2024.0984

Adapted from a piece published on the Cambridge Judge Business School website.

The text in this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Images, including our videos, are Copyright ©University of Cambridge and licensors/contributors as identified. All rights reserved. We make our image and video content available in a number of ways - on our main website under its Terms and conditions, and on a range of channels including social media that permit your use and sharing of our content under their respective Terms.