In 2004, the AMI Consortium set out to collect a multimodal meeting corpus that would give us all the raw material we needed to demonstrate a whole range of meeting support technologies, most of which we knew we hadn't thought of yet. I will talk about how we designed the corpus to grow an interdisciplinary community that would collectively understand not just the technologies but how groups work, and then I will describe some of the novel applications we have built using the data and are currently showing to industrial end users.