The development of urban water system solutions has long relied on manual workflows dating back to the 1950s, which depend heavily on expert knowledge for task decomposition, model building, and decision-making – significantly limiting efficiency and scalability. Advances in large language models (LLMs) are unlocking new possibilities for automating complex workflows. This perspective explores the transformative potential of LLM-based workflow automation in the development of planning and management interventions for urban water systems. A multiagent framework is proposed, featuring an orchestrating artificial intelligence (AI) agent for workflow decomposition and specialized agents for tasks such as environmental perception, data analytics, modeling, optimization, and solution evaluation. Balancing AI and human decision-making is identified as a critical challenge, necessitating tailored approaches to address the complexities of specific problems and dynamic environments. Key barriers to the adoption of AI agents are analyzed across six dimensions: ecosystem and infrastructure, data, ethics, technical limitations, institutional challenges, and social implications. This perspective envisions AI agents as pivotal in enhancing the efficiency, productivity, and scalability of current manual workflows, thereby driving innovation in water system planning and management.
Guangtao Fu (Wed,) studied this question.