Taking vast quantities of spend data, from multiple sources, in varying degrees of completeness, and then turning it into a single, coherent whole is no mean feat. The bulk of the heavy lifting in that process is handled by the patented AI engine of GEP SMARTTM. Supported by a complex program of human oversight, industry expertise and a multi-billion record data set, GEP SMART’s AI constantly learns and retrains, looking for exceptions and data that “doesn’t fit”.
Using whatever clues are available in your raw spend data, GEP SMART can examine every line of every invoice, looking at every currency unit of spend, and then assign a probability of a certain classification to that spend. If the classification is not obvious enough, GEP SMART looks for matches, classification rules and other hints from the invoice data to see what else can be used to classify the spend.
Simply put, if you can look at the invoice and — with some investigation — classify the spend correctly, then so can GEP SMART. For example, if an invoice from a supplier has only the total amount but that supplier only provides you with steel bearings, then there's very high probability that steel bearings is the correct classification. However, if the supplier sells bearings to one of your business units but provides engineering consultancy to another, a straightforward examination of the business unit to which the invoice was sent can help classify the spend correctly.
With many more layers of complexity and sophistication, this is how GEP SMART’s AI can learn and build an accurate model of your spend.