Two years ago, three economists at the World Bank -- Michael Lokshin, Zurab Sajaia and Sergiy Radyakin -- cooked up ADePT. That, to my knowledge, was the first successful attempt to use Stata for leveraging econometric research across the world.
Think of it this way: a star academic economist builds a theoretical model that explains the extent to which pro-growth policies are also pro-poor. This model can be used for tracking the effectiveness of poverty reduction efforts across the world, but for that you need to pair the econometrics with the data collection in some practical way: people in far away places upload data and receive the estimation results, complete with warnings about outliers, missing observations, etc.
ADePT does two very valuable things very well. One is that it packages the most expensive input, the academic's time and effort, and replicates it at practically zero marginal cost. The other is that it cuts out a lot of drudgery. Without it, periodic reporting at the World Bank would consist of collecting and cleaning the data, estimating a model and outputting formatted output as many times as needed, by hand. This would be done with statistical software designed for doing this only once, by people whose opportunity cost of time does not justify doing this more than once.
Well, ADePT is the shape of things to come. Soon you will able to do web-based automated data analysis, for private-sector clients, using Numerics by Stata, an embedded version of the Stata computation kernel. You read it here first.