Can you envision sophisticated robots driving the front and back-end operations for banks? While the answer would have been uncertain a few years ago, today is an emphatic yes. Most of the capabilities needed to realize a digitally driven retail bank are derived predominantly from Artificial Intelligence (AI) and Machine Learning (ML). These technologies are hardly new, but they are gaining traction. AI investments in global banking will grow approximately 33% yearly through 2030, reaching an estimated $64 billion.
Companies follow relative market trends and customer demands to help determine future investments. Many trends (Figure 1) demand investments in AI and ML. The banking use cases mentioned below require analyzing large datasets in real time. The good news is that AI/ML offers high-speed processing power, the ability to join data and provides real-time data access. These features make it easy to achieve hyper-automation goals when adding AI to the mix.
Figure 1: Trends in the banking industry
AI-based Cognitive Computing for Banking
Opportunities to use AI for cognitive computing in the banking sector can be divided broadly into three parts – cognitive engagement, cognitive automation and cognitive insights. Cognitive engagement is improving the user experience (UX). It is a core tenet that provides seamless user engagement during customer-facing activities. Cognitive automation uses AI-driven robots and Robotic Process Automation (RPA) to enable back-office hyper-automation of various mundane tasks like image and text processing for form filling and straight-through processing (STP) of other repetitive tasks. Cognitive insight looks into multiple sources of structured, unstructured and numerical data to build actionable insights for the bank. It requires processing huge volumes of disparate data and using AI to enable near real-time fraud detection and regulatory compliance. These opportunities allow banks to leverage AI in many STP tasks and high-volume banking processes.
AI in Retail Banking: Four Core Use Cases
Deploying AI solutions can get tricky. Banks can partner with IT firms to leverage in-house cognitive assets and an AI partner ecosystem to enable core use cases like account opening, customer support, fraud detection and credit risk scoring.
a. Account Opening
The current process, in most banks, is still time and effort intensive. A combination of AI-driven hyper-automated solutions and RPA can make the process effortless for customers as they only need to visit the bank’s website and upload the soft copy of necessary documents for account opening. The automated process will involve multiple system integrations and the use of AI/ML for text/email classification, OCR, ICR, automated voice calls, text-to-speech, speech-to-text, process automation, and more. Infusion of AI, as explained above, leads to enhanced user experience, reduction in manual efforts and mundane tasks, and swift turnaround time.
b. Customer Support
Support is another pain point for retail banking customers, especially when placing a service request outside the bank’s working hours. AI-driven hyper-automation can provide twin benefits — enhancing the user experience and streamlining back office operations. Customers can call and converse with an intelligent voice-enabled bot 24/7 to report a lost card, lodge complaints, and book an appointment with a bank department or other predefined service. Another valuable use case for AI is analyzing user actions in near real-time and responding to such events proactively. An example is monitoring a dubious credit card transaction and alerting the user to take appropriate action.
c. Fraud Detection
Sophisticated fraud is rising, even as banks adopt solutions to detect and mitigate fraudulent transactions. For banks to confirm a fraudulent transaction, most solutions depend on data models to find out-of-the-ordinary transactions to be reviewed by a bank employee. These solutions are prone to human error and consume resources and time.
An AI-based anomaly detection engine can report a fraudulent bank transaction and trigger an Intelligent Interactive Voice Bot to call the customer and take feedback. If the customer acknowledges the transaction, the bot records a false positive, and the system learns from this interaction. Alternatively, if the customer denies making the transaction, the bot marks it as fraud, blocks the card, and initiates a new card request for the customer.
d. Probability to Default
Probability to default is a financial concept where models calculate the chances of a borrower defaulting on a credit repayment. The model can speculate if and how the customer will default on the monthly card payments and calculate the total amount owed and the loss to the lender.
Multiple factors affect the borrower’s ability to repay a loan – overall economic condition, geography, industry structure, type of collateral, etc. A risk model is needed to ingest historical records from a CRM to learn the default pattern, ingest external data (inflation, interest rates, GDP, etc.), and correlate with default probability. This risk model leverages AI and advanced machine learning approaches to provide insight into the borrower’s ability to service the loan as a risk score. Moreover, multiple regulations mandate that such AI models should be explainable and not a black box. It should allow an SME to look at the supporting evidence for the default predictions. Multiple cloud AI/ML workbenches have already started providing “Explainable AI,” leveraged for loan approvals or rejections. The latest trend in AI/ML tools and techniques allows all the features above to be easily added and help predict the default probabilities.
Banking on AI to Remain ‘Always-On’
Amid shifting customer demands and increasing competition from incumbent and disruptive new entrants, banks realize they must be available to their customers – regardless of the time, device, or channel. Whether it’s core banking operations or advanced security needs, customers expect access to the bank anytime, anywhere. Leveraging a combination of AI and RPA can help banks do just that by hyper-automating processes to improve efficiency, enable quicker response at less cost, and ensure service consistency. After years of incremental changes in silos, it’s time for banks to adopt AI and fundamentally reimagine core operations to suit the new digital reality. Case in point:
a) The Wipro AI solutions team helped a leading bank in North America digitize their contract documents which involved OCR, automatic document classification, attribute extraction and a Smart user interface. This solution led to benefits like a reduction in manual effort and productivity improvements.
b) A UK-based bank was migrating to Google cloud. The bank built its advanced analytics and machine learning applications using proprietary software (SAS) and hosted them on-premise. Wipro helped the bank transfer these applications to the Google Cloud Platform using its core-flex approach resulting in license-cost savings.
John Loney
Global Head, AIOps Alliances & Product CoE, Wipro Limited
John has a background in building products and automating processes in enterprises using AI and cloud-based solutions. At Wipro, he focuses on building the AIOps and Automation ecosystem and works with partners to drive pipeline and joint GTM. John’s team creates accelerators and reusable components that complement partner products and help with fast and efficient delivery.
Manish Okhade
Practice Head, Cloud AI, Wipro Limited
Manish has over two decades of industry experience in several domains like ER&D, Digital and Artificial Intelligence. He is presently leading a Cloud AI Practice in Wipro’s AI/ML Solutions group. Manish’s current focus areas are building accelerators, thought leadership and forging partnerships with niche start-ups in various domains.