Description of Problem:

There were various time-consuming processes in the Operations Department of Client’s company. Most of the processes were rules-based, while some required machine-learning (an aspect of artificial intelligence).

1. Booking of containers when client’s end-customers place orders
   • This process required staff from Client’s Company to log on to government’s web portals to make bookings and declarations.
2. Reports of booking orders/ incoming cargo with detailed information about containers
   • This process required staff from Client’s Company to log on to government’s web portals to get data about cargo.
   • The information had to formatted in a specific way.
3. Transferring information from Bills of Lading (Excel) into desired format and preparing a Manifest, which is a combination of various Bills of Lading
4. Extracting information from Bills of Lading (scans and PDFs)

How Gleematic software helps:

1. Booking of containers when client’s end-customers place orders
   • Gleematic was able to complete the whole process of form-filling onto two government web-portals.
2. Reports of booking orders/ incoming cargo with detailed information about containers
   • Gleematic was activated upon departure/arrival of vessels to get information from two government portals.
   • Information from web-portals were retrieved quickly by clicking through various tabs and storing in a temporary memory.
   • The information was then placed into an Excel file and formatted in a specific way to be sent to end-customers.
3. Transferring information from Bills of Lading (Excel) into desired format and preparing a Manifest
   • Gleematic was able to transfer data from one Excel file to another of a different format with near-perfect accuracy to create a Manifest
   • The data in the Manifest was formatted by Gleematic so that the information fits into one page.
4. Extracting information from Bills of Lading (scans and PDFs)
   • Gleematic was trained with about 90 different copies of Bills of Lading (BLs) of a specific format by tagging various fields of data. The purpose was to build a “model” to read this format of BLs.
   • Logic and coding were included to break up strings of text into distinct categories.
   • The “model” was then tested on 10 new copies of BLs of “Harbout Link Lines” format and achieved accuracy of about 85%.

Challenges:

1. Web portals were sometimes slow to load and run into errors at times. Various “exception shipping handling” scripts were set up to address different types of scenarios. For example, Gleematic robots were set to look out for a certain image to ascertain if the page has loaded. In another scenario, a sub-script was activated when Gleematic encounters a specific error message.
2. Formatting data to fit nicely onto an A4 page for printing was challenging, as text sometimes flowed to the next page. It was finally done after multiple ways of testing. If there is not sufficient for setting up Gleematic robot, this step can be easily done by humans manually.
3. Machine-learning to extracting information from Bills of Lading required multiple sets of data of the same format. Accuracy can be further enhanced by re-training the Gleematic robot with another 100 copies or more of BLs.

Benefits:

• Shorter Time of completion: Gleematic cognitive automation works up to 5 times faster than humans.
• High Accuracy: Almost 100% accurate. There would be no human-errors of missing data, uploading wrongly, or having duplicated files in the archival process.
• FTE Saving: ~3 to 4 FTEs
• Saves precious hours of humans: As the human-users would only have to monitor the robot’s progress, he/she can spend most of his/her time on other more important things.
• Reduced stress on staff: As the staff (humans) need not attend to this mundane and repetitive yet detailed work, they would have less stress and be more motivated to do other value-added jobs.