AI may be making companies less creative, new study warns

The AI tools that companies increasingly rely on for innovation may actually be narrowing their thinking, according to new research published in the Academy of Management Journal.

The study finds that mainstream search engines and large language models (LLMs) are delivering the most popular and familiar answers, creating clusters of similar ideas, or 'ideation bubbles', that quietly suppress creativity.

“Managers are asking teams to think outside the box while giving them tools designed to keep them inside it,” said Hila Lifshitz, Professor of Management at Warwick Business School, one of the study’s authors. 

“If organisations want breakthrough ideas, they need systems built for exploration, not just efficiency.”

A design problem - not a talent problem

The researchers argue the issue is structural, not human. Most digital tools operate on what is known as 'exploitation' logic - ranking mainstream, high-probability results first. However, breakthrough thinking requires 'exploration': exposure to diverse, unconventional inputs.

“Most companies use search and AI tools that are optimised for efficiency,” said Moran Lazar, Assistant Professor of Entrepreneurship and Innovation at Tel Aviv University.

“That’s great for execution, but problematic for innovation, because it pushes everyone towards the same ideas and creates what we call ideation bubbles.”

To test whether this could be fixed, the team built an exploration-based layer on top of standard search technology (Google Search). A simple, smart algorithmic layer called 'XYZ' - named after the 'long tail' of innovation where creative ideas often live - deliberately surfaced conceptually diverse results instead of clustering similar ones at the top of the page.

The approach was tested in two rigorous environments:

• A lab experiment with 104 participants
• A global innovation challenge on Freelancer.com involving 245 participants from more than 40 countries, focused on reducing household food waste.

Participants using the exploration-based tool generated twice as many distinct idea clusters as those using a standard search, and overall creativity scores rose significantly. In short the experiment found that changing the algorithm changed the thinking.

The study also challenged the assumption that AI mainly helps novices perform like experts. In fact, experts benefited most from exposure to diverse inputs. While experienced professionals can become entrenched in established patterns, they are uniquely capable of connecting distant ideas when prompted with unconventional material.

“The key takeaway for leaders is that in the age of algorithmic disruption, innovation is not just about using algorithms to produce more ideas, it’s about designing tools and roles so that human-machine collaboration is optimised for innovation,” said Professor Lifshitz.  

“Otherwise, organisations will gain an efficiency boost from algorithms, but not an innovation one.”

What leaders can do now

The researchers say organisations can act immediately by: 

• Asking AI for ideas from other industries or diverse viewpoints
• Requesting multiple alternative framings instead of one ‘perfect’ answer
• Increasing AI ‘temperature’ settings to prioritise variance over precision
• Tracking ‘ideation bubbles’ by clustering and monitoring idea dominance patterns.

Ultimately, the research shows that boosting innovation isn’t about generating more ideas with AI, but about redesigning the algorithms themselves so they expand, not narrow, the creative possibilities available to people.

• This article is based on the following research: Lazar, M. Lifshitz, H., Ayoubi C. and Emuna, H (2025). Would Archimedes Shout “Eureka” with Algorithms? The Hidden Hand of Algorithmic Design in Idea Generation, the Creation of Ideation Bubbles, and How Experts Can Burst Them Academy of Management Journal. doi: 10.5465/amj.2023.1307
• This is a summary of their research Breaking Out of Bubbles Created by Innovating with Algorithms. Academy of Management Journal Insights. doi: https://journals.aom.org/doi/full/10.5465/amj.2023.1307.summary 

Further reading:

Can AI predict people's future illnesses?

Understanding the essence of leadership today

Three quick wins for managers using generative AI

Is bias causing business leaders to make mistakes?

Hila Lifshitz is Professor of Management and Director of the Artificial Intelligence Innovation Network. She is also a visiting faculty at Harvard University's Lab for Innovation Science. She teaches Digital Transformation on the Executive MBA and Global Online MBA, and Managing and Leading Digital Innovation on the MSc Management of Information Systems & Digital Innovation.

Follow Hila Lifshitz on Twitter @H_DigInnovation.

Learn more about digital innovation and AI on the two-day Executive Education course AI Leadership at WBS in London | Warwick Business School.

Discover more about the Future of Work and Digital Innovation and Entrepreneurship. Receive our Core Insights newsletter via email or LinkedIn.