When does AI become worth it? A multi-attribute utility perspective for urban climate planning

Urban climate modeling is essential for resilient city planning but remains constrained by high computational costs and expert dependence. We introduce KLIMASCANNER, an AI-based QGIS plugin for rapid microclimate predictions, and employ Multi-Attribute Utility Theory (MAUT) to compare it against conventional case-by-case (CBC) modeling across five dimensions: energy efficiency, speed, transparency, cost, and accuracy. Monte Carlo analysis (N=10,000) with three uncertainty sources shows KLIMASCANNER is preferred in 94–99.9% of scenarios depending on stakeholder profile, with break-even at approximately 40 simulations. Sensitivity analyses confirm robustness to cost assumptions and weight configurations, though CBC remains competitive when transparency is prioritized. AI-assisted modeling can democratize urban climate assessment for smaller municipalities, while numerical outputs alone cannot fully capture decision-making complexity in climate-resilient planning. • KLIMASCANNER is an AI-based QGIS plugin that predicts how urban developments affect local temperature, wind fields, cold air corridors, and thermal comfort indices at 5 m resolution. • Multi-Attribute Utility Theory (MAUT) with Monte Carlo uncertainty analysis (N=10,000) provides a rigorous framework to compare AI-based and physics-based climate modeling approaches. • KLIMASCANNER is preferred in 94–99.9% of simulated scenarios, depending on stakeholder profile, with citizens showing the highest uncertainty due to transparency considerations. • Break-even analysis indicates AI-based tools become cost-effective after approximately 40 simulations, democratizing climate assessment for smaller municipalities.