Leverage AI to predict whether each application should be approved, rejected, or manually reviewed and even manual review team (branch, region, head office)
Configure different levels of automation or AI assistance for different segments, sectors, products and credit levels
Guide credit origination teams to most important factors influencing credit level definition and automation
Loan default prediction
Challenges
Many credit decisioning systems are driven by scorecards where these scorecards had been simplistic rule-based systems in the past. In the recent years, organisations started to develop scorecards with the traditional techniques. However, developing scorecards, maintaining them based on the macroeconomic dynamics and behavioral needs are difficult and time consuming
Opportunities
Determine lending risk and payment default rates in situations using different level of information
Simplify the process of predicting whether the customer will default on their credit or not to
Minimize credit losses which will an important input for the provisions of the bank
Grow healthy portfolio limits of the institution
Explain the reasoning behind the scoring thanks to Machine Learning Explainability algorithms (Shap values)
Handset limit assignment
Challenges
Improper and incorrect income estimation
Short limits for affluent clients
High limits for low-income customers
Lack of automation for the models and using outdated models
Opportunities
Utilize AI algorithms to increase the power of income estimation models and use 360-view of the customers to define the final income of the customer
Assign limits based on the income of the customer which will allow accurate and proper limits for each customer
Automate continuos-targeted models which will ensure inflation and other macroeconomic factors will be considered in the new models
Pricing
Challenges
One-fits-all pricing for each customer
Lack of competition in the market for pricing
Loss of price-sensitive customers due to pricing
Loss of revenue for non-sensitive clients
Opportunities
Develop pricing models with the best-in-class AI algorithms
Differentiate pricing for each segment, product and customer
Increase through-the-door customer flow with the campaign prices
Increase revenue with the right pricing
Early warning prediction
Challenges
Lack of definition for the customers who tend to miss their payments in the near future
Standard actions for whole customers independent of customer's risk status
Opportunities
Determine customers with the payment stress
Differentiate early warning actions for stressful customers
Improve credit roll rates and portfolio quality
Decrease operational action costs and potential future collection costs
Collection
Challenges
Determine customers with the payment stress
Differentiate early warning actions for stressful customers
Improve credit roll rates and portfolio quality
Decrease operational action costs and potential future collection costs
Opportunities
Machine learning models to estimate forward roll rate probability
Increased collection rates
Decreased operational costs
Better provisioning in the company level
Churn prediction
Challenges
Loss of customers without prior notice
No action for the customers based on their leaving probability or value to the company
Opportunities
Identify customers who tend to leave in the near future with the AI algorithms
Take actions to retain valuable existing customers based on churn prediction models
Maintain revenue from the existing portfolio
Propensity prediction
Challenges
Limited number of products and limited revenue from the customers
No proactive cross-sell or up-sell actions
One-fits-all product campaigns
Opportunities
Identify customers who tend to buy additional products or increase usage in their existing products
Take proactive cross-sell or up-sell actions to get share in the market for specific products from the valuable customers
Grow the portfolio and product coverage
Activation prediction
Challenges
Non-activated products after product asssignment to the customers
Lack of usage of the products
Opportunities
Increase in the active customer base
Increase in revenue from the products
Segmentation
Challenges
Lack of granular customer segments
Traditional customer segmentation without analytical methods
Opportunities
Granular customer segmentation including risk segmentation or marketing segmentation concepts utilizing advanced clustering AI algorithms
Customized customer management or risk activities based on customer segments
Customer lifetime value estimation
Challenges
Limited information regarding the value of the client
Standard actions to approach each client independent of the value of the client
Opportunities
AI models can help calculate the lifetime value of the clients
Increased revenue
Customized client actions for affluent clients
Fraud
Challenges
Lack of strong fraud models
Lack of adequate number of fraud events in the data
Opportunities
AI is ideal for anomaly detection which helps to capture unusual events in the behaviour of the customer
Fraud data events can be increased within the portfolio sample with oversampling, undersampling or class weight features