Agentic AI Researcher: Initial Experiences and Outlook

22nd International Conference on Network and Service ManagementAlcalá de Henares (Madrid), Spain // 26 - 30 October, 2026 Keynote by Christian Esteve Rothenberg, University of Campinas (UNICAMP), Brazil What happens when an AI system does the research itself? For this year's CNSM Agentic AI Researcher track, we gave an AI researcher one prompt and let it write a complete paper: the research question, method, simulation code, experiments, figures and text, with no human edits. The paper was accepted, and three AI reviewers scored it 78, 99 and 95. This keynote shares what we learned from that experiment. It covers what the AI did well, where it fell short, what the reviews reveal about using AI as reviewers, and who should get credit when the key idea comes from the human's prompt. It also draws on lessons from ongoing AI-assisted research practices and on my own work building a "digital brain" that gives AI assistants lasting memory and context. The talk ends with an outlook and some open questions for the network and service management community. https://www.cnsm-conf.org/2026/keynotes.html