I recently read an article at OnJava: "Esper: Event Stream Processing and Correlation". Event Stream Processing is a very cool technology that allows real-time reaction to events in your business. It can be used effectively in network monitoring and business intelligence.
Rather than reacting to receiving a single event, which doesn't scale well from a network monitoring perspective, you react to an aggregated view of those events. If one request to your database is slow, that isn't significant in a large application. If you look at the request times over a 1 minute period, and see that the latency is increasing; that's very significant. Stream processing allows you to process these sort of events.
I'm not a business guy, so any example I come up with for that is going to be totally contrived and lame. I'm sure there are some great ones though :P
Have a look at Wikipedia for more explanation.
Back to the article...
The article focuses on using an open-source tool caled Esper, which is avaialable from Codehaus. I haven't set the application up yet, but I will on one of my two 40 minute train rides. We use a similar product where I work. It is not open source, and is not available to me here. Politics :)