Water systems experience increasing sustainability challenges from climate variability, aging infrastructure, and energy and chemical intensity demands, but AI has typically been assessed against prediction accuracy rather than demonstrated operational success. This PRISMA 2020 systematic review analyzed the role of AI solutions on sustainability in distribution, treatment, and basin management. The database search identified 920 records; after deduplication (n = 185), screening was conducted on n = 735 titles/abstracts and examination of the full text for n = 85, providing a total of n = 41 included peer-reviewed studies for qualitative synthesis and n = 38 for quantitative/bibliometric synthesis with the additional analysis of seven grey-literature sources. Evidence mapping reveals high growth post-2020, and distribution and wastewater operations are dominated by a few companies. The most deployable evidence is found with monitoring, anomaly/leak detection, and short-term forecasting, while optimization and reinforcement-learning control are primarily simulation validated with limited field applications. While accuracy metrics are often reported, transformation into water saved, kWh/m3, chemicals, compliance/reliability/resilience/equity measures are inconsistently and less frequently operationalized. In general, AI is most believable when it is part of analysis-ready workflows, bounded decision support, and measurement-and-verification.
Jorge Alejandro Silva (Mon,) studied this question.