Decreasing Manual Workload by Automating SAP Travel Expense Workflows

Abstract

In the 21st century, efficiency is a key focus for several organisations, and because of this, machine learning and process automation is getting a lot of attention. The Norwegian Government Agency of Financial Management deals with a large amount of travel expense claims every year, which makes them a possible point of interest for process automation. The claims are at this point approved by manual labour, but we will research the possibility of automating parts of this process by using historical data extracted from the SAP backend system used. To perform this automation, several machine learning methods will be tested to perform a classification on the data. As there are attachments involved in a lot of the claims, Optical Character Recognition will be used to perform a processing of these. We failed to produce a solution that could perform good classification on the extracted data, but our results prove that there it is possible to solve this problem in an optimal manner.