RPA. AI Implementation at Multinational IT Provider
IT Service Provider
Our customer is a Japanese multinational provider of information technology services and products. Headquartered in Tokyo, Japan, the company provides IT and network solutions to business enterprises, communication service providers, and government agencies.
Software & IT
The customer aimed to test feasibility of a cognitive automation technology for its business processes. The pilot business process consisted in validation of the tax amount charged in a vendor contract. Tax rates vary for different groups of vendors and are defined in the customer’s database.
- Achieve overall business process accuracy of 98 percent
- Reduce manual time spent on one process by 50 percent
- Train and support customer’s team to build automation for an SAP case.
- Create an information extraction model for the fields vendor_name and cost, implement a predefined data storage with a list of vendors, and conduct an additional similarity check. Create a stamp detection model based on Tensorflow, Keras, and Open CV. Develop a script that locates a stamp in a document to help the business analytics team prepare a machine learning training set with 100 percent correctly marked data.
- Create five RPA nodes for MS Access automation using a desktop driver. As the database is encrypted, JDBC options are not applicable.
- Create a single RPA node for Skype automation with infinite reading of incoming messages to trigger the launch of the pilot business process.
- Develop an automatic tool to decrypt documents online.
IBA Group acted as a subcontractor, working on the customer’s premises. The project was ongoing within 1.5 months.
IBA Group and the customer organized regular Q&A sessions to overcome language and cultural differences, transfer knowledge of RPA/AI technologies, drive SAP case development, and resolve issues with software and source code.
As the databases were encrypted in line with the customer’s policies, IBA Group developed an automatic tool that decrypts documents online using InfoCage FileSystem Shell.
- Overall business process accuracy: 100%
- Reduction of manual effort: by 83.6%
- Average document processing time: 30 seconds
- Cognitive automation trained on 1,469 documents or 4,000 pages.
Ultimately, the customer’s development team built their own automation for an SAP case.