In smart spaces such as smart office and workplaces, users are surrounded by hundreds of networked devices and services which often vanish into the background. Therefore, the users are often unaware of possible tasks which they can achieve within a given physical space. Moreover, the user does not want to dig in a long list of individual services and devices for functionalities which are required to accomplish a task at hand. To enable doing more with less in everyday life, we envision a future intelligent computing where the users should not handle directly functionalities provided by individual services and devices but rather high-level tasks, e.g. ‘watch a movie’ or ‘borrow a book’. In this paper, we present our vision of a task computing framework which is to deal with development, deployment, discovery, recommendation, and execution of high-level tasks within smart spaces. We outline a research agenda aiming to realise the proposed framework.