Agentic AI and generative large language models are rapidly reshaping software engineering practices. In requirements engineering, a discipline grounded in language, negotiation, interpretation, and context sensitivity, their impact is particularly profound. Requirements are textual, social, and iterative artifacts. They emerge from conversations, trade-offs, and evolving stakeholder needs. As AI systems become capable of participating in interviews, drafting specifications, analysing feedback from stakeholders, or supporting strategic analyses, the boundaries of traditional RE activities begin to blur. However, much current research focuses on adapting isolated RE tasks to LLMs as black-box components. Are we optimizing the wrong unit of analysis? Is the paradigm shifting from task automation to lifecycle redesign? What does it mean to evaluate not just outputs, but processes, collaboration dynamics, scalability, and sustainability? The talk highlights the need to move beyond single-task benchmarks toward holistic, end-to-end perspectives on the RE lifecycle, to reconsider evaluation and dataset design, and to reflect on how AI agents should be positioned within RE teams.

I am an assistant professor at the Department of Service and Information System Engineering (ESSI) and a member of the Software and Service Engineering Research Group (GESSI), from Universitat Politècnica de Catalunya (UPC-BarcelonaTech). My research is focused on the field of requirements engineering, natural language processing, knowledge graphs and AI-based systems.