With rapid technological progress, there is growing demand for efficient methods to evaluate and forecast future technologies. This study proposes a novel Delphi survey methodology using AI-generated expert agents powered by large language models (LLMs). These agents simulate interdisciplinary expert panels and engage in iterative discussions to assess the developmental stages of emerging technologies, with improved reliability through integration with external knowledge sources. This study addresses three research questions: (1) Can AI-generated agents accurately assess technologies predicted over a decade ago by contextualizing historical predictions against present-day reality? (2) Can they reliably evaluate recently predicted emerging technologies with consistent evaluation methods across temporal contexts? (3) Can they forecast next-generation technologies by synthesizing technological trajectories and identifying innovation opportunities? Experiments were conducted using two bio-technology lists from 2012 and 2021, respectively. Results indicate that AI agents accurately recognized older technologies as commercialized and newer ones as still in development. Furthermore, the agents identified next-generation technologies, such as in vivo gene editing and AI-enhanced multimodal imaging. These findings demonstrate the potential of AI-generated expert agents to support scalable, cost-effective technology foresight. Retraining these agents on previously evaluated technologies may further enhance their predictive capabilities, enabling more strategic alignment with future innovation. • AI-generated expert agents with diverse personas successfully conduct multi-round Delphi surveys for technology evaluation. • AI agents in Delphi surveys demonstrate accurate tracking of technology trajectories. • RAG integration improves URL validity and increases evaluation consistency by 68%. • Model differences inform use: Claude for speed and GPT-4o for in-depth RAG analysis. • Agents identify 12 next-gen biotechs, aiding R&D investment and strategic planning.
Lee et al. (Tue,) studied this question.