Economist and R blogger JD Long gave a talk last week (as part of the vconf.org project) about why he uses R to do statistical forecasts of agricultural yield for the reinsurance company he works for. I couldn't make the live session, but a replay is now available. The audio's a bit choppy, but if you've every struggled with using Excel for doing anything more than the simplest analysis of data and (in JD's words) you're suffering from "Excel Exceedance Syndrome", it's worth it for the first five minutes alone. With a regression example (the code and data are available) shows how he not only gets great analysis out of R, it's also reproducible and faster than using other methods. JD also does a great job of turning a technical problem (lack of the right files on the demo system) into a serendipitous demonstration of the flexibility of R, by downloading the extra data and packages on the fy. There's also some good discussion on making use of Amazon EC2 when you need extra power from R, too.
vconf.org: R the Language