Banks and financial institutions (BFSI) continue to invest heavily in intelligent automation technologies to meet the ever-increasing demand for speed and personalized service from tech-savvy corporate banking clients while complying with the barrage of compliance and regulatory restrictions that govern them. Intelligent automation technologies combine the power of robotics process automation (RPA) with artificial intelligence (AI) technologies like optical/intelligent character recognition (OCR/ICR), machine learning (ML), natural language processing (NLP), and conversational platforms to solve repetitive and relatively complex human operations and interactions with speed and consistency in service.
We’ve helped our BFSI clients benefit from intelligent automation in their operations. Here’s a snapshot of the efficiency we’ve delivered:
- Commercial lending operations:
- The commercial lending lifecycle typically starts with onboarding a corporate client. ML, OCR, and RPA were used to extract data from structured and unstructured loan documents and automate the entry of client data into the account opening and loan processing systems and calculation of risk and loan eligibility. The overall loan processing lifecycle was accelerated by over 30%.
- Trade finance operations: The lifecycle consumes significant operations bandwidth due to multiple and disparate systems and high levels of documentation.
- Letter of credit (LC) and guarantees: Efficiencies of over 20% in the LC processing and LC advising process were achieved by training ML engines to extract data from LC applications and using RPA to enter details of the applicant and beneficiary, amount, and other details; calculate charges; and work with the SWIFT system to create an MT700 message that was sent to the advising bank.
- Bill discounting and financing: Requests for financing and discounting was automatically processed through an ML engine trained in multiple formats and reduced manual interventions by over 15%. Multiple steps in the supply chain and working capital process was also automated.
- Import and export payments: The combination of OCR, ML, and RPA was used to process multiple documents such as cover letters, bills of lading, airway bills, certificates, and invoices and enter data into trade finance systems to process payments for importers. Intelligent automation was also used to streamline export payment operations by over 20% and reduce manual interventions to reconcile incoming payments to exporters with the trade finance and accounting systems.
- Regulatory and compliance operations: There are multiple regulatory and compliance requirements in corporate banking that lead to high operational effort.
- Know your customer (KYC), anti-money laundering (AML), and sanction screening: The names of people, companies, and other entities found in client documents were intelligently extracted and fed by RPA to sanction screening systems with ML used to identify false positives in name screening results; this process significantly reduced the cost of compliance operation. RPA was used to pick up KYC candidates and automate the entry of data into AML applications like LexisNexis and process the results. More than 25% of effort in client onboarding operations was reduced with KYC and sanction screening automation.
- Transaction screening: Banks have large operational teams working on managing sanction hits with multiple levels of approvals before clearing a false positive, but more than 95% of sanction hits are false positives. We trained an ML engine on the most common patterns of false positives to automatically remove them, leave an audit trail of every bot decision for compliance reviews, and significantly reduce operational effort.
- Payment operations: Intelligent automation was applied to automate the processing of incoming and outgoing payments, reduce manual effort to investigate payment issues, increase straight-through processing rates for payments by over 40%, and reduce the dependency on operators.
- Inward/outward payments: Intelligent automation was used to automate multiple payment processing scenarios to extract relevant data from payment reference fields in SWIFT MT202 messages and accelerate NOSTRO transfers, extract details from fund transfer forms for incoming payments, automatically manage payroll processing and vendor payments, and initiate outgoing fund transfers.
- Payment investigations and repair: ML was used to identify common patterns of payment repairs and automatically fix payment messages that were not automatically processed due to incorrect or missing account numbers, BIC codes, or correspondent bank details. The investigation of frequent payment processing issues such as a beneficiary claiming non-receipt of funds or incorrect account details was automated with ML and RPA to process unstructured content in MT199 messages, send responses, and read from and enter data into the payment systems. RPA was further leveraged to handle issues like payment recalls and exchange rate discrepancies by applying a set of rules.
- Cash management operation:
- A lot of labor is required to process invoices and purchase orders and reconcile them with the bank’s accounting systems. Because of the high volume of invoices, the bank’s checkers were sampling and verifying only about 10% of invoices before approving them. We trained the ML engine to identify multiple invoice format structures and extract the relevant data. Invoices that were not recognized were sent to an operator to process with the ML engine learning in the background with RPA to enter data into accounting systems. This reduced operational costs by over 30% and allowed the bank to sample 100% of invoices consistently.
There are many, many more instances of intelligent automation helping banks and financial institutions stay competitive in this marketplace of evolving customer expectations and growing regulatory compliance.