Abstract
Generative AI has ignited a mixture of excitement and fear within the foresight community. It is therefore important to explore the implications of widespread AI adoption for the future of our field. AI serves as an exceptional research assistant to expedite the horizon scanning process, as it can sift through vast volumes of data much faster than a team possibly can. As part of a horizon scanning process a systematic inventory of key drivers forms the basis of many Trend Scenarios in foresight studies. This study investigated whether the application of a more systematic approach for trend identification using AI could be used to collect more reproducible results than standard methods. Moreover, we specifically examined whether AI could help to better distinguish between types of drivers, long-term trends and (early) signals that are still too weak for a trend, but could have an impact. Using CHATRIVM key publications were scanned with specific prompts and results were compared to an overview of trends using commonly used methods for trend identification, such as a (grey) literature search and stakeholder consultation. Given the objectivity of trends and megatrends, we've found that AI can project their future trajectories with considerable accuracy. Yet, AI's analysis of trends and key drivers is typically standard and may not yield novel insights for those well-versed in the subject. It's useful as a starting point, but results must always be validated and refined by human analysts.