A New Low Risk Approach to AI for Consumer Lending


AI and machine learning have enormous potential, particularly in the competitive consumer lending space. But there are frustrating impediments to its development, deployment and use that are slowing down value realization in many companies. The requisite talent is hard to find, attract, retain and manage. Big up-front investments in data platforms and tools may be necessary. And even with the talent and the data infrastructure, there are no guarantees of success. Some executives, facing such uncertainties, might be forgiven for concluding that AI is only for the largest and most sophisticated players or for the fintech specialists. SymphonyAI’s aim is to change all that with their EurekaAI targeted AI powered solutions for business people.

While machine learning techniques are powerful, knowing how to use them and where to apply them remains the foremost obstacle. Unfortunately, data scientists, IT and business people often do not understand each other. Business people know what they need, data scientists know what is possible and IT people know the company’s systems, but putting all the pieces together can seem all but impossible.

Common challenges include:

  • Knowing how to define the problem, which data to use, and how to use the results. Even the best data science teams if, as is all too common, they have limited business experience and are thin on banking product expertise can be stymied by these basic questions
  • Business leaders may not fully understand what is possible with advanced analytics, and might be intimidated by data science jargon
  • Even when successfully developed, getting analytics “the last mile” from the lab into the field can be tricky, requiring collaboration across data science, data engineers, IT, and business
  • Finally, getting full adoption and trust from business users can be a challenge if the new capability is not communicated well or people are not fully trained on its use, purpose, interpretation and limitations

SymphonyAI has taken a new approach with their EurekaAI suite of solutions. Instead of AI for data scientists, EurekaAI solutions put analytics driven insights directly into the hands of business analysts and decision makers. EurekaAI makes AI intuitive, so that informed decisions can be made with confidence, insight driven actions can be taken, and business value can be realized quickly.

Making this possible begins with coupling world class data science technology with deep business expertise. The award winning EurekaAI platform is the foundation for analytical solutions across SymphonyAI’s vertical businesses in healthcare, retail, life sciences, financial services, media and public sector. EurekaAI is an end-to-end machine learning platform that automates many of the complex steps required to build predictive models such as feature selection, model selection, model tuning, model validation and model deployment. EurekaAI combines powerful machine learning capabilities with their industry leading “topographical data analysis” to both reveal insights into the shape of the data and to help make predictive models more accurate.

EurekaAI allows both data scientists and “power users” in the business, or “citizen data scientists,” to explore complex datasets and build powerful predictive models without the coding in R or Python that is typically necessary. But “autoML” solutions (that automate the machine learning lifecycle) still require skilled practitioners to tease out the most valuable use cases, those that will generate meaningful business performance improvements. This is where EurekaAI solutions comes in.

With EurekaAI solutions, SymphonyAI does that work for you. In their latest solution, SymphonyAI has identified the highest value uses of AI in auto lending through extensive market research in consumer auto finance covering the full lifecycle: from prospecting and client acquisition, underwriting and approval, to pricing strategy, collections and loss mitigation. They then prepackaged these use cases in a powerful and intuitive user interface in their EurekaAI for Consumer Lending solution aimed at business people.

Through its research, SymphonyAI identified the highest value opportunities to apply AI in auto finance for banks, credit unions, fintechs, and captives, and then built those use cases on it’s EurekaAI platform. The result is a rapidly implementable, low risk solution to help business leaders:

  • Find the best clients, with the most attractive (risk adjusted) profit potential,
  • Assess delinquency risk more accurately than competitors,
  • Measure price sensitivity across the credit spectrum,
  • Spot market pricing inefficiencies and opportunities, and
  • Optimize collections effectiveness and efficiency

EurekaAI for Consumer Lending learns from your own data to determine which borrower characteristics best predict delinquency risk, to predict borrower price elasticity, and to optimize collections. With data driven insight you can optimize your origination, underwriting, pricing, portfolio management, and collections strategies, bringing performance improvements to both your top line and your bottom line.

Here are three examples from the EurekaAI for Consumer Lending solution for auto lenders.

The “Credit Insights” module provides insight into portfolio composition and performance using standard metrics such as credit score and delinquency, while also providing clues to what drives client behavior using sophisticated machine learning techniques. This can reveal which borrower characteristics are predictive of higher or lower delinquency and default risk.

Credit Insights

Business people (product managers, pricing managers and risk managers) can use this simple yet powerful tool to understand how their portfolio is performing, and with sophisticated underlying AI capabilities, to identify hidden data features which are predictive of borrower behavior. And this can be done at granular portfolio segments such as vintage, credit score at origination, LTV at origination, term, location, and even (for auto lending) the make and model of vehicle.

In a second example, EurekaAI for Consumer Lending uses past data on loan offers made and loans booked to provide actionable insight into borrower price sensitivity in combination with the likelihood of delinquency. Using sophisticated models, built transparently and automatically, a business person can see how a small change in price can affect the probability of acceptance. Product managers can determine optimal pricing strategy to get more of the business they want, or to see how higher rates will impact volume. This is done through AI models that learn from past client behavior to predict the probability that a client will accept an offer at a particular price. These models are built automatically requiring no data science expertise on the part of the user.

Business users can interactively model how different pricing strategies will affect production volume across different borrower or product segments. A product manager might use these insights to design a portfolio that balances new loan volume, margins, and collections workload. In certain segments, delinquency data may indicate lower risk. There product managers can predict where a big increase in volume is possible through a small decrease in price. And in segments where your data shows your delinquency risk to be higher, product managers can identify segments where the volume falloff from an increase in price is likely to be minimal.

Price Elasticity and Missed Opportunity

In the bottom half of this screen is an innovative capability called “Missed Opportunities.” The premise is that there may be desirable loans you are not making, or you are missing because pricing is too conservative. By comparing known delinquency behavior with similar loan applicants EurekaAI identifies borrowers likely to behave similarly. This may reveal, for example, applicants that may have been mis-scored who actually have low delinquency risk. By using these two capabilities in tandem, a product or risk manager might adjust underwriting criteria or pricing to miss fewer profitable opportunities.

For many consumer credit companies, collections is the single highest (non-interest) expense component. Correctly prioritizing borrowers is the key to increasing both collections effectiveness and efficiency. EurekaAI learns from your historical delinquency data to predict who is likely to become late, to go from late to delinquent, or to self-cure. With this insight you know how to prioritize workloads for collections staff.

Using a visually intuitive “Sankey diagram,” EurekaAI depicts in one visualization both how delinquencies have changed in the past, and how delinquencies are predicted to change in the future. Interactive capabilities allow business users to explore which late borrowers are likely to become delinquent and which are likely to self-cure so collections staff can focus their efforts where needed most.

Delinquency: Actual and Predicted

Other capabilities within the EurekaAI Consumer Lending solution include prospect insights which help you analyze your funnel of applicants, offers and loans booked, customer acquisition to help you build effective marketing campaigns, and loss mitigation to help you determine the best strategy for severely delinquent or defaulted loans.

Already, SymphonyAI is working on a module with use cases customized for home equity credit products and plans additional modules for credit card issuers and small business lenders.

AI is hard. To apply, to develop, to deploy and to use AI effectively takes depth in the business, data and analytics. But SymphonyAI has made it it’s mission to take the complexity out of AI. Using business expertise and a world class AI platform, they are building AI for business people.

Up until now, building and developing AI solutions required a potentially significant upfront investment in people, platforms, and tools. And there was no guarantee of success. EurekaAI offers AI powered solutions that business leaders can deploy and use to drive business performance without the high expense and execution risk of “do it yourself” AI solutions.