Traditionally, distributed computing concentrates on computation understood at the level of information exchange and sets aside human and organizational concerns as largely to be handled in an ad hoc manner. Increasingly, however, distributed applications involve multiple loci of autonomy. Research in multiagent systems (MAS) addresses autonomy by drawing on concepts and techniques from artificial intelligence. However, MAS research generally lacks an adequate understanding of modern distributed computing. In this Blue Sky paper, we envision decentralized multiagent systems as a way to place decentralized intelligence in distributed computing, specifically, by supporting computation at the level of social meanings. We motivate our proposals for research in the context of the Internet of Things (IoT), which has become a major thrust in distributed computing. From the IoT's representative applications, we abstract out the major challenges of relevance to decentralized intelligence. These include the heterogeneity of IoT components; asynchronous and delay-tolerant communication and decoupled enactment; and multiple stakeholders with subtle requirements for governance, incorporating resource usage, cooperation, and privacy. The IoT yields high-impact problems that require solutions that go beyond traditional ways of thinking. We conclude with highlights of some possible research directions in decentralized MAS, including programming models; interaction-oriented software engineering; and what we term enlightened governance.