Digital Lending: How Artificial Intelligence, Machine Learning Are Making a Difference

machine learning digital lending

Artificial intelligence (AI) and machine learning (ML) are ubiquitous in today’s workplace conversations. Turn on any business news channel and you’ll hear them repeated over and over. Ask any venture capitalist and they are sure to brag about several investments in these areas. Google artificial intelligence and machine learning, and you’ll find 213,000,000 hits, and […]

The post Digital Lending: How Artificial Intelligence, Machine Learning Are Making a Difference appeared first on Lending Times.

machine learning digital lending

Artificial intelligence (AI) and machine learning (ML) are ubiquitous in today’s workplace conversations. Turn on any business news channel and you’ll hear them repeated over and over. Ask any venture capitalist and they are sure to brag about several investments in these areas. Google artificial intelligence and machine learning, and you’ll find 213,000,000 hits, and rising. Overhyped? We don’t think so.

Accenture boldly claimed that AI could boost average profitability rates by 38% and lead to an economic benefit of $14 trillion by 2035. That is no small statement. Even more astonishing is the general alignment among analysts on this issue. It’s widely agreed that AI and ML hold great promise across all industries, and, specifically, in finance.

In 2019, IDC projected that banking would be the second largest global industry to invest in AI, with $5.6 billion going toward AI-enabled solutions (trailing only retail). Why? The anticipated effect on business. According to the research firm, Autonomous, the financial industry’s slice of the global AI pie represents upwards of $1 trillion in projected cost savings.

Fintech Disruptors and Underwriting

Fintech disruptors, characterized as fast-moving companies, often start-ups, focus on a particular web-based innovative financial technology or process, spanning mobile payments to lending. Fintech disruptors initially found an entry point in finance through the use of AI/ML in underwriting.

In the U.S., if the customer consents, you can gain almost unlimited data about their credit profile: how many loans they have, whether they have a mortgage, if they’re delinquent, and whether they requested credit recently. According to the Brookings Institution, “AI coupled with ML and big data, allows for far larger types of data to be factored into a credit calculation. Examples range from social media profiles, to what type of computer you are using, to what you wear, and where you buy your clothes.” Access to this type of data gave rise to the development of sophisticated algorithms to underwrite consumer credit risk. We’ve seen this across a variety of lending companies offering unsecured consumer, student, or even small business loans, particularly focused on digital lending.

Importantly, though, those employing AI must be hyperaware of data collection practices, model design, and the potential for misuse. There is an inherent obligation when using these powerful tools to avoid profit at any cost. When used responsibly, AI can promote growth and better serve consumers. To meet this goal, companies must focus on creating ecosystems that are exponentially more just and equitable than what we have today.

On the surface, the digital lending numbers seem incredible. Digital lenders have grown to $50 billion in originations per year, not including incumbents. And, the research firm Autonomous notes that the digital lender model continues to raise $5 billion in annual venture capital investment, dominated by investments in the U.S.

And, yet, that same report shows that an AI/ML-driven digitization of the lending process is not headed to zero cost. To date, the cost advantages of onboarding and ongoing servicing (up to 70% reductions) have not been able to overcome the relatively high marketing costs that have yet to effectively scale lower than $250 per loan. Moreover, capital costs can reduce efficacy relative to traditional bank competition, and, then, there are the unplanned expenses, such as legal fees or elevated product development costs, the firm reports.

So, if digital lending driven by AI/ML-powered underwriting cannot deliver a material cost advantage, is further AI/ML advancement possible? And, will it improve outcomes for the consumer? Yes, absolutely. It all boils down to operations. As the use of AI shifts beyond obvious use cases and is deployed cross-functionally across entire companies to address various operational inefficiencies, the real promise emerges.

AI/ML 2.0: Improving Outcomes for Everyone

According to Deloitte, the top 30% of financial services firms who are frontrunners are more adept at integrating AI into the core strategic business of their firms, delivering revenue and cost gains quicker than competitors. In our opinion, this is clearly the case with fintech disruptors. Those that are focused on AI integration throughout the organization will quickly pull ahead of those who limit AI deployments to chatbots, underwriting, and other AI/ML 1.0 use cases.

Fintech disruptors can offer the market’s most cost-effective solutions by dramatically curtailing operation costs. Harnessing large-scale, multi-functional AI systems across organizations, instead of simply deploying in underwriting, presents fintech disruptors the opportunity to control costs at each stage and offer quality outcomes for their customers at reduced costs – with lean workforces.

So, while these systems may not face the end customer in any way – in fact, that may not be visible at all – they are the true future of AI/ML for fintech disruptors.

Fintech disruptor leaders who understand the opportunity to use an interconnected system of AI models across their organizations will likely drive the greatest overall efficiencies, both reducing costs and boosting revenues. This enhanced efficiency can be used to drive competitive position and ultimately higher profits.

AI/ML 2.0 at Work

AI can be used to help allocate resources across a variety of functions. For instance, a lender could create an AI model used to predict which of its retail partners would see the greatest increase in usage as a result of a field visit by a partner support representative. Generally, these visits don’t have uniform outcomes. Therefore, using a model-driven approach could help to allocate resources in the most effective manner. Increasing usage obviously drives overall revenue, but also helps to amortize cost over a greater number of transactions, driving better unit economics. Further, with time, the usefulness of such a system can grow. The more data collected from previous visits, the better the algorithm can be at predicting which visits will yield increasing usage.

Or, a lender could deploy AI in the call center to optimize the efficiency of the collections support team. Outbound reach to delinquent customers could be prioritized based on an ML algorithm that evaluates the potential for a successful call and the expected dollar collection. This may sound simple, but making the “good” calls and avoiding the “bad” ones offers all the obvious advantages of more precise resource allocation.

What is less obvious, though, is how these models are interconnected. The model used in the call center complements the underwriting model. If the collections team performs better, then the underwriting model can be recalibrated to maintain the overall risk of the loan portfolio. If the model prioritizing field visits is working, then it increases usage and reduces the average costs to originate a loan. This further enables a recalibration of both the underwriting model and the collections model. The combination of these models, ultimately, increases both expected and realized returns on the loan portfolio, reducing expenses and allowing the company to pass this savings back to customers in the form of lower rates. This is a win for everyone.

Optimizing the AI/ML Ecosystem

This is the true promise of AI/ML – a robust ecosystem of interdependent models utilized to enhance cross-functional outcomes. This leads to a much broader point: inefficiencies exist in all aspects of business – including accounting, legal, operations, finance and customer experience – and negatively impact profits.

Responsibly managed AI/ML 2.0 promises to address many of these functional silos with great success, improving outcomes for everyone involved.

Author:

Dr. Tamir Hazan is a co-founder and head of Analytics at Digital Lending: How Artificial Intelligence, Machine Learning Are Making a Difference appeared first on Lending Times.