A fascinating article in Wired, Dec. 2005 at 280 explains how data entered into a computer can help art painting experts detect forgeries. Analogously, computer analysis of texts can help prove authorship, such as the several Federalist Papers that are not directly attributable to Hamilton or Madison. Similarly, a company has analyzed popular songs according to many characteristics and can identify and recommend related music if given songs someone likes.
Each of these tools uses software, large amounts of data, analytic algorithms, neural network concepts, and taxonomies of characteristics. The tools aid experts who want a quantified, deep, and objective foil for their somewhat intuitive deductions.
I thought of invoice analysis. A law department accumulates a pile of long and elaborate bills from a law firm. With the same approach, why could software not detect subtle differences in management techniques (firm’s and law department’s), productivity of timekeepers, assignments of staff, milestone achievements and other patterned actions too hard to notice by manual review?