Renewable energy systems (RES) in the residential sector, like photovoltaic systems, heat pumps and battery storage, are corner¬stones of a sustainable energy supply. Nevertheless–and despite major fiscal stimuli–private investment in such technologies has not yet reached a satisfactory level, also because sale of such products is time-consuming and requires a high level of expertise from suppliers. In practice, small and medium-sized installation firms are often responsible for addressing customers, advising, designing and implementing the appropriate systems, but they struggle with offering the complex technology and are exposed to fierce competition in their market. In a joint research initiative with a RES supplier and a software development company, we drive the development of information systems that support installation companies in their tasks. To this end, we are using action design to develop a cognitive computing solution based on Machine Learning (ML) to promote the sale of sustainable energy products. Based on 4,909 real customer requests for RES and survey data from 666 homeowners (which we use as ground truth data for ML), a predictive model can reliably identify promising RES installations out of a list of customer requests and thereby supports an important business task. Despite these promising results, we face a number of challenges in developing our cognitive computing solution. To address these challenges, design principles for similar systems are developed, contributing to the current debate on how information systems research can support sustainable development and how artificial intelligence can be used profitably in enterprises.