Generative AI is able to produce dozens of ideas in response to a problem in the blink of an eye. While the potential capacity to enhance our creativity may appear considerable, the results of early experiments are more tempered.

A study conducted by Stanford University and the consulting firm GeoLab shows that we still have some way to go. The researchers asked several teams to generate solutions to a real-life problem by drawing on a pool of data. The time allotted was 90 minutes; only some part of the teams had access to ChatGPT. The generated ideas were then graded from A (very convincing) to D (of no interest) by the concerned company’s decision-makers.

Surprisingly, the use of generative AI only very marginally increased the number of ideas generated (+8%). Although it did help reduce the number of bad ideas (–7%), it also slightly diminished the number of excellent ideas (–2%). Indeed, generative AI tools are designed to deliver the most statistically probable answers. They therefore tend towards the average, rather than towards truly innovative ideas.

It is therefore best to start with a reflection process that is not influenced by AI-generated ideas. AI then excels at helping to iteratively improve these ideas. Learning to interact with AI to deepen your ideas will thus become a key competency.

Source: Don’t Let Gen AI Limit Your Team’s Creativity, Harvard Business Review, March-April 2024.