Unleashing the power of Artificial Intelligence (AI) and Machine Learning (ML) in Banking sector

by Gaurav Bansal - Business Consultant –Digital Transformation
| minute read

What used to be just a dream in the era of science fiction, Artificial Intelligence (AI) is now mainstream technology and is impacting almost everyone’s lives directly or indirectly, with applications in image and voice recognition, language translations, chatbots, and predictive data analysis.

So how do Machine Learning and Deep Learning work? To understand in the simplest possible manner, Machine Learning uses algorithms to parse data, learn from that data, and make informed decisions based on what it has learned. On the other hand, Deep Learning, structures algorithms in layers to create an "artificial neural network” that can learn and make intelligent decisions on its own.

To make it work you feed the model with huge amounts of sample data to “train” it. With enough data samples, the model can accurately “predict” output for any future input. For example, if you load a machine learning program with a huge dataset of x-ray pictures along with a relevant description (diagnosis, symptoms, and other information), it will learn from the dataset and would be capable to assist (or perhaps automatize) the data analysis of x-ray in the future.

In this article, I will discuss a few of the solutions of AI and how these can be leveraged in the Banking Domain. 

Understanding the need of AI/ML in the Banking sector

Digital disruption is redefining industries and changing the way businesses function. Every industry is assessing options and adopting ways to create value in this technology-driven world:

  • With the introduction of open banking and PSD directive, there are huge possibilities of data exchange between financial institutions.
  • Some of the highest paid jobs are in the financial sector, so if a Bank can reduce a small amount of workforce by utilizing AI, there can be huge cost savings for the Bank.
  • Banks have a humongous amount of structured/unstructured customer-related data that can be leveraged to train models.
  • With the increase in the digital ecosystem, consumers are using online mediums for financial transactions and Banks are opting for Branchless banking. 

AI/ML landscape in Banking

Fraud Detection

Operations Automation

Customer Experience

Credit Decisioning and Automation

Artificial Transactions

Conversational search for internal repository

Virtual Assistant/Bots

Historical  portfolio analysis

Duplicate invoicing

Queue Optimization (In branch)

Targeted Advertisement

Financial information analysis

Credit card fraud

Automated Document processing 

Customer churn

Digital footprint analysis

ATM Fraud

Optimizing support using historical tickets 

Auto KYC

Credit report analysis

Signature fraud

Risk Reporting

Behavioral analytics

Post Disbursal Monitoring

AML Identification

Live chat

Spend Analytics

Third-party Data analysis

The above table highlights a few of the prominent applications that leverage AI/ML within banking.

For ease of understanding, we have divided these applications into four service areas and a few of the use cases from each area are explained below:

Fraud Detection

Machine Learning and Deep Learning can help existing data analytics applications recognize potential fraud cases while avoiding acceptable deviations from the normal. In the case of wrong alerts, these deviations can be flagged and marked as false positive, thereby aiding in fine-tuning the application.

  • Artificial Transactions – Machine Learning can be used to identify patterns to detect circular transactions (between two businesses that belong to the same customer) or inflated transactions (used to temporarily increase the cash flow).
  • Anti-Money Laundering (AML) –  Machine Learning coupled with Deep Learning can help in reducing false positives for sanctions thereby improving the quality of alerts.
  • ATM Frauds – Computer vision and video analytics can be used to identify a person fraudulently using ATM.
  • Credit Card Frauds - Anomaly detection and spending patterns can be used to increase the accuracy of credit card fraud detection.

Customer Experience and Engagement

With the increase in competition, every bank needs to retain its customer by providing an enhanced customer experience. There can be multiple ways to enhance customer experiences, some of them are by reducing service time, providing personalized services, round the clock support, and continual engagement. AI/ML can be leveraged to build applications to improve customer delight.

  • Virtual Assistant/Bots -  Virtual assistants provide support 24/7 without human efforts, leveraging Natural Language Processing (NLP) to understand the user’s intent and respond to basic queries.
  • Auto KYC – To extract information from user documents,  Machine Learning and NLP helps to identify information such as Age, Address, DOB, etc. and to detect fraud by identifying Digitally manipulated images.
  • Targeted Advertisement - Applications can use customer transactional data, to identify cross-selling activities, and create personalized campaigns to target customers.
  • Customer Behavior Analytics - AI can help analyze behavior such as how many times a customer has visited the branch, frequency of transactions, lifetime value, how many times a customer has sought support in the past, customer feedback, etc. Such insights on customer behavior can be used for planning and strategy.

Operations/Back Office Automation

The consumer-facing applications of Machine Learning (ML) are receiving most of the attention today. However, with increased pressure on banks from regulators to comply with regulations and increase in competition from Fintech firms with new business models, it has become a necessity for banks to leverage AI technologies for streamlining and automating their operations.

  • Conversational Search for the internal repository - Ontologies based repository can be created for internal documents, natural language processing along with other AI techniques can be used to have a Google-like conversational search engine to find out relevant content/document from the repository.
  • Queue optimization (In-Branch) - Digital queue can be created in branches, IOT enabled tokens or video analytics can be used to optimize the service time for the customer and provide analytics & insight for effective resource planning at branches.
  • Document processing - Cognitive OCR along with Machine Learning and Deep Learning can be used for the extraction and processing of information from the document.
  • Optimizing application support - With the help of Machine Learning, banks can improve application support by auto-predicting priorities, resolutions, and identifying the best support staff to resolve the tickets quickly.

Credit Decisions and Monitoring

Traditional credit scoring methods are becoming redundant - as the world is changing, so is the scoring methodology. Instead of relying on just the credit score (as in the traditional approach), creditors are looking towards AI based credit scoring applications that help underwriters analyze multiple data sets along with their linkages during credit assessment. Some examples are given below :

  • Historic portfolio analysis - Learn from past data to unlock hidden patterns that can be used in making credit decisions. The model can be used to create a credit score, which can take into consideration multiple data points such as customer clusters, demography, obligations, age, etc.
  • Digital Foot Print analysis - Derive value from the digital footprint of an individual from sources like social media posts and Internet activity, what sites someone visits, what they purchase from eCommerce stores, utility bill payments, etc. Online behavior can help in predicting whether a person is likely to pay back their loans or not.
  • Post Disbursal monitoring - Providing an accurate assessment of the borrower’s risk level post disbursement is important, AI enabled continual monitoring of loans can help banks in taking corrective actions in case of probable defaults. Banks can plan their collection strategy in case of likely defaults, effectively engaging with the customer by giving incentives for early payments, etc.
  • Financial information analysis - Process Financial statements, Bank statements, Credit card statements to identify hidden information, patterns, links such as cash flow trend, fixed obligations, cheque bounces, overdrafts, validate profit and loss statements, balance sheets, etc. Intelligence can also be built to identify reasons for the decrease in profits or increase in interest payments from annexures and notes.

Our Offerings

At Sopra Steria, we continuously thrive on exploring new opportunities within Artificial Intelligence and Machine Learning. We have several home-grown assets that can be leveraged to accelerate the customized development of any of the applications described above.

The future of banking lies in AI/ML and it would be remiss for any bank to not leverage it.  With Banking 2.0, we are seeing the rise of Branchless banking. It is quite possible that we experience efficient Bot-driven banking in the near future with minimal dependency on human resources.