Thanks @Ozzie Gooen and @Aaron Gertler, for the detailed comments. As Ozzie pointed out, the findings in collective intelligence (CI) are indeed very scattered, in the sense that what we know about CI exists as bits of effects that are often not connected. While real-world scenarios are typically affected by many interacting social and psychological factors, and how these effects interact is oftentimes unclear. I don't think there is an up-to-date synthesizing effort in the field at the moment, which is my frustration too, and what I'm trying to do in my work, and it's no easy task). As a young transdisciplinary field, CI still has much work to do in organizing and synthesizing. In fact, they just started to have their own journal (Collective Intelligence published by ACM) very recently. In the area of organizing, I think CI can learn a lot from the EA efforts. 

Good question on the applications of collective intelligence findings, and of the role of technology. I don't have in-depth knowledge about these areas. However, a resource that could be helpful is Handbook of Collective Intelligence (2005), edited by Bernstein and Malone. It devotes a chapter to discuss organizational behavior (effects on teams and organizations), and another on AI. Also see Nesta on some efforts applying collective intelligence research findings in policymaking and beyond. 

As for collective intelligence in the more complex scenarios, there is a study on project groups, another on solving a large engineering problem, also a Huff Post article on CI for "mega problem-solving." The work done on complex, real-world problem solving is not much compared to those done in labs. I hope there to be more on this in the future. Start-ups would be a great subject to study! I haven't seen any work on this, though.