Nowadays, bookkeeping must deal with a significant amount of documents that must be processed on a monthly basis. This process involves both the need to enter the data from documents into the systems and account it accordingly, which is not only tedious and time-consuming but also prone to manual errors.
Datarabbit, a member of the PrediqTank consortium, is working with e-file, one of Poland’s top companies providing accounting software on workflow for automizing invoice processing, with the overall goal of providing a solution capable of making it error-free, as well as requiring less effort and actual work from the accountants.
In particular, our team uses the latest image-processing neural network advancements in order to implement an OCR system, that automatically reads all text from the invoices (electronic documents or photos), and later extracts from it the information which is actually crucial from an accounting perspective, significantly diminishing the need for manual data entering. Later the pipeline is also capable of automatically alerting about anomalies/uncommon situations that may need special treatment (such as unusual tax rates). Our work also included deployment and integration of the developed solution with the client’s existing Azure cloud environment. This needed to be done in a scalable fashion, capable of reacting to a variable workload.
"I am impressed of the knowledge and ideas of Datarabit team and their OCR product. You can use OCR as an all-in-one tool that is custom-tailored, affordable, and automize costly processes.”
~ Sławomir Nowak, e-file senior software developer
