The challenge
Every day, hundreds of order confirmations arrive in Van Marcke's mailbox. Usually, this information is embedded in a pdf file, from which information has to be copied into Van Marcke's mainframe. Until recently, this process was done entirely manually. After realizing this was an extremely time-consuming and non-value-adding task, Van Marcke called in the expertise of our sister company RoboRana, as they are specialized in hyperautomation (and Robotic Process Automation). However, RPA wasn't enough to fix this problem. Therefore, they called in our AI expertise.
Together with RoboRana, we combined artificial intelligence and RPA into a single solution. Our AI tool, Klassif.ai, receives the PDFs from a Blue Prism RPA bot. During the next step, the tool uses optical character recognition (OCR) technology to convert the unstructured data, such as product names, quantities, and amounts, into structured data.
To recognize the unstructured data, the AI tool had to be trained first. In the early stages, the employees had to intervene manually by showing the tool where to find an amount or indicating the difference between an invoice and delivery data.
Through this kind of training, the system became smarter and gains insight into how the orders are usually structured. As a result, the tool can now classify Van Marcke's order confirmation without help.
As a final step, one of RoboRana's virtual workers enters that structured data into the mainframe. As a result, the order confirmation process happens fully automatically instead of manually.
The time and efficiency gains from this intelligent process automation story are enormous. We're talking about several hours per day saved.
Moreover, the error margin that was involved in manual processing has also been reduced, as both the interpretation by Klassif.ai and entry by the RPA bots are as good as error-free. Employee happiness has increased as there is more time to do interesting tasks instead of repetitive ones.
"In the first month, almost 7.000 PDFs were processed. If you know that an employee usually needs 1 to 2 minutes per document, this results in 10 hours saved per day on average. This is an immense amount of time that can be used for more value-adding tasks."
