There are two main problems with BI software.
First, too little thought is put into the post go-live analysis portion of implementing a BI package. All the effort is spent setting up software, gathering reams of data, and cooking up a few canned reports, and at the end of the day, when you’re sitting atop a mountain of data everyone looks around and says “now what?” More thought should be put into what the key drivers of that particular business are (there should be a handful, not a pile), and providing “real people” (i.e. not from IT) the tools and training to mine this pile of data as they see fit.
The second problem is with business reporting as a whole, and the reason why many BI implementations are billed as unsuccessful. By its very nature, BI looks at the past, and it is easy to look at your stack of fancy reports and believe it contains the tools you need to predict the future direction of the business. This is abjectly false, and the best example is the financial industry. Finance always has the highest IT spend in the industry, and a very high penetration of BI tools, yet nearly all the big financial organizations missed the recent financial crisis. You need strong management that realized BI provides clues to the future, but not the future itself. IT often is a partner in crime, overselling this aspect of BI.
Like all software, BI is a tool to be leveraged, it is not a magic potion that will predict the future, despite all the fancy dashboards, reams of data and fancy analytical terms.