International, integrated container logistics company - the world’s largest shipping company with largest container line with 17% market share. In the world of transportation and more specifically cross-border freight traffic and customs clearing, there is a long and cumbersome process. Often this process involves a lot of documents and is managed manually.
The process includes collecting information from truck and train release documents. Typically, only 60% of the documents are in a standard format. These documents contain customs declarations for each truck or train and the goods they transport in each of the containers - due to disruption of received documents, goods information must be adjusted multiple times in Helios system to subject to the Governal Customs administration.
However, Helios fails to enable advanced changes and his nature complicate adjustments and prolongs these cases. The documents are scanned pictures and typically, an original copy or the digital version of the text is not available. Therefore, all information about each container must be rewritten manually into the customer’s CRM system. On top of that, documents may contain multiple declarations divided in sections of which only some are relevant for the process. Within the PDF documents these sections can start anywhere throughout the document since the structure is not predefined and can reflect random circumstance in rail freight traffic.
The customer used to divide the documents by printing and scanning the parts relevant to specific truck and train declaration which requires resources and higher processing time – this became inaccessible during the pandemic crisis when the office was not used as the prime workplace and professional equipment was not available to the employees. Since documents are scanned pictures, the quality differs, which leads to deciphering number and letter combination which is time-consuming and demanding for human sight.
We developed two robotic processes that help with
Result
Employees do not need to read and search for related information from the scanned documentation, rewriting shipping codes and manually adding each piece of information into freight system. Reading hundreds of codes for each transport vehicle, correctly rewriting those took hours of tedious, eye staining work. In case of special case, employees are notified to do an extra control to mitigate risk of error. This creates time and space for employees to dedicate this time elsewhere.
Implemented in
4
weeks
Time savings
120
hours/month
Accuracy
93%
ROI
3
months
For the operation in the Czech Republic, we use trains and trucks to transport the containers and there is a lot of manual work involved in handling the process. Robot ICT helped us automate 2 significant processes using UiPath, which involved handling a variety of different documents and internal and external applications to declare goods with custom authorities. The projects were delivered in time, handover of the services was done very professionally including training. Everything was well documented as part of the handover and till now they continue to support us making sure the Robots can run 24/7. We are very satisfied with the company's skills and approach.
Milan Jakubec
Operations Manager CZ, SK
Instead of having to perform basic activities like rewriting numbers from documents to a CRM system
The software robot is precise and performs checks on extracted data.
Return of investment was 3 months.
Today's commonly used OCR - Optical Character Recognition technology can easily "read" text from any scanned document. Depending on what data we need to extract, regular expressions (RegEx) can be used to find, for example, codes with a known format. Another option is to train a machine learning model - ML. A certain amount of documents is manually annotated (hundreds to tens of thousands) and the mathematical model can then recognize the given fields on other, new documents. The more variable the structure of the documents, the more time-consuming the learning of the model and the required amount of training data.
Depending on the type of document, we can compare the extracted data with other data. This means, for example, a checksum on a document or a partial comparison with another accompanying document. We can also compare the data against some codebook or database.
RPA Consultant & Python Developer