Prescriptive technology intelligence for technology opportunity discovery: An LLM-based automated framework for narrating promising technology concepts

Existing technology intelligence approaches have largely remained at descriptive or predictive levels, limiting their capacity to provide specific and actionable prescriptions for technology opportunity discovery. This study introduces the concept of prescriptive technology intelligence and proposes a corresponding methodological framework. We present an LLM-based end-to-end framework that automatically generates narrative technology concepts from technological documents with minimal human intervention. Specifically, the framework constructs a technology knowledge graph and expands it by predicting latent relations and identifying structural technology themes. These components are then integrated and converted into explicit technology concepts using a graph-to-text model, with an LLM-as-a-judge approach further screening and ranking the most promising opportunities. An empirical demonstration using a large-scale dataset of artificial intelligence conference abstracts confirms that the proposed approach successfully transforms raw technological data into actionable R&D directions, thereby effectively facilitating strategic technology planning.