Recommender systems assist customers to make decisions; however, the modest adoption of digital technology in the water industry means no such system exists for household water users. Such a system for the water industry would suggest to consumers the most effective ways to conserve water based on their historical data from smart water meters. The advantage for water utilities in metropolitan areas is in managing demand, such as low pressure during peak hours or water shortages during drought. For customers, effective recommendations could save them money. This paper presents a novel vision of a recommender system prototype and discusses the benefits both for the consumers and the water utility companies. The success of this type of system would depend on the ability to anticipate the time of the next major water use so as to make useful, timely recommendations. Hence, the prototype is based on a long short-term memory (LSTM) neural network that predicts significant water consumption events (i.e., showers, baths, irrigation, etc.) for 83 households. The preliminary results show that LSTM is a useful method of prediction with an average root mean square error (RMSE) of 0.403. The analysis also provides indications of the scope of further research required for developing a commercially successful recommender system.