This study integrates Large Language Models (LLMs) with the Theory of Inventive Problem-Solving (TRIZ) to enhance systematic innovation in engineering. We compare traditional TRIZ with an AI-assisted approach in a case study aimed at reducing water consumption in cooling towers of a urea plant in the petrochemical industry. Using structured prompts, the LLM generated over 10 solutions, evaluated for novelty and feasibility by human experts. The AI-assisted method reduced analysis time from one week (traditional TRIZ) to four hours, achieving a 90% efficiency gain. Key solutions include multi-stage hybrid systems and optimized airflow, potentially saving 10–20% water based on preliminary computational fluid dynamics (CFD) simulations. We discuss generalizability to other domains, LLM limitations such as hallucinations, and the need for human validation. This framework offers a scalable, time-efficient tool for addressing water-intensive engineering challenges.