I recently commented on a post in a business intelligence forum, and I wanted to share that comment here for others to consider. The forum was discussing the challenges of implementing business intelligence and analytics:
One of the concerns I have with automating business intelligence is that many problems that I am confronted with from my clients are simply not "tame" (in other words, problems we understand, and in which ample data is available). Solving "tame" problems is much different than solving "wicked" problems (in which we have little understanding of the variables or conceptual framework, and in which little pertinent or valid data is available). Clearly, "tame" problems should be automated as much as possible. However, "wicked" problems are often the "writ large" of business intelligence requirements. Not all problems can or should be automated as the first step in solving the problem at hand. Also, automated business intelligence must be transparent enough that validation and reliability testing can be performed manually. Business intelligence is a vital growth area for enterprise in the 21st century, but we must still take care to do the job right in accordance with best research practices.