Our imperative was to exceed the technologies that banks currently use for performing two critical tasks: adverse news analysis and transactions screening.
Working on the leading edge of
ML technology, ITRex helped deliver two sophisticated AI-powered RPA tools combining advanced
NLP,
AI, and data processing capabilities to autonomously perform these tasks.
AI-powered RPA bot for adverse news analysis
This RPA bot for banks efficiently handles the initial level of screening news for flagged individuals and entities that can be potentially involved in sanctionable actions. It is designed to augment existing capabilities in customer onboarding, KYC, and financial crimes. Its key features are:
●
Automated Media Review: Leveraging data parsing techniques, the RPA bot uses identifiers such as the name, year of birth, and country of origin of the subject to search news aggregation platforms for article analysis.
●
Customizable Search Parameters: Users can choose from a range of platforms (like Google, TR, Factiva), set specific keywords, and limit article numbers, customizing the tool to meet their needs.
●
Intelligent Article Selection: Unlike standard rule-based bots, the system employs advanced ML algorithms with integrated NLP capabilities to determine whether an alert is a false positive or a legitimate concern, focusing on negative mentions. The ML model is trained with high-quality sector-specific data for enhanced accuracy and is capable of providing feedback for further review.
●
Keyword Highlighting and Relevance Assessment: The RPA bot highlights mentions of the subject and specific keywords within articles to speed up manual review.
●
Decision-making: Based on decisions made at every step, the system decides whether an article raises a legitimate concern for teams to intervene.
●
ChatGPT Integration: It offers the option to use ChatGPT via OpenAI’s API for analyzing articles after they have been normalized. ChatGPT returns structured summaries for the system to decide whether the piece warrants further investigation.
●
Data Storage: The system stores metadata, input data, and records of additional manual verifications, ensuring a comprehensive audit trail for all decisions made during the review process.
AI-powered RPA for transactions screening
This RPA bot for the banking industry specializes in the automated screening of transaction sanctions alerts, crucial in the current era of heightened global financial scrutiny, particularly with recent geopolitical events and the rise of real-time payments (RTP).
●
Diverse Format Processing: Capable of processing a wide range of payment messages, including SWIFT, ACH, and ISO 20022, as well as custom message types, using advanced NLP for parsing and semantic analysis.
●
AI-Powered Decision-Making: Uses sophisticated ML algorithms to autonomously recognize and match names for people, organizations, countries, and other entities in payment messages, comparing them against a comprehensive decision matrix and historical transaction records.
●
Real-time Processing: Capable of screening real-time payments for sanctions hits instantly, ensuring uninterrupted payment flows.
●
Accuracy: Achieves a remarkable 70%+ reduction in false positive alerts.
●
Decision Explanability: Delivers written justifications for each screening decision and is also capable of consulting third-party sources for additional validation. When transaction legitimacy remains uncertain, the system escalates these cases for further scrutiny.
●
Broad System Integration: Features built-in support for major compliance systems
