Leveraging the power of AI to enhance Payment Investigation solution for a Global Finanical Group

Challenges.

The customer had a sizable volume of Payment Investigation messages that came in daily through the incoming channels. All these are free-format and open length text messages. Payment Operations had to read through each message and identify the appropriate information from these messages, both structured and unstructured, before updating the Payment workflow. 

The Payment Investigation process had inherent challenges due to a lack of standardization, data accuracy and completeness, fraudulent transactions, legacy environment etc. While integrating new technologies is costly and time consuming, addressing these requires a combination of process improvements and technology upgrades. Introducing point solutions for process automation by using AI enhanced accuracy, efficiency and was an effective way of getting quick ROI. 

The overall objective was to enhance the efficiency of the payment investigation process within the bank’s back-office operations. Specifically, the initiative aimed to automate the manual step in the process by implementing an AI solution. The solution was to bring consistency, scalability, and accuracy to the Payment Investigation process 

Solution.

This system was designed to accurately extract key fields from six commonly used, unstructured SWIFT message types (MT195, MT196, MT199, MT295, MT296, and MT299) involved in payment investigations. Structured SWIFT (MT/MX) as well as unstructured Payment messages are consumed by a Kofax Total Agility (KTA) bot. The AI model developed used Named Entity Recognition (NER) and Multi-Class ML classification to extract information from identified fields. For a specific text field, a Multi classification ML model with SVC (Support Vector Classification) algorithm was used. Word Cloud was created to understand keyword distribution across each Reason type. The existing KTA workflow was updated to integrate the AI solution and facilitate the current human in the loop for verification of extracted values. Extracted information was updated in the KTA workflow along with confidence factor and Information displayed to investigator for validation 

Design 

• Open source based (OSS) solution was cost effective and flexible to meet evolving needs of the Bank.
• Seamless integration design to integrate with upstream and downstream applications.
• Scalable architecture to keep up with the increase in demand.
• It is customizable for onboarding, enables new use-cases and features with minimal effort 

The impact.