How AI is going to change lending
Despite all the buzz about artificial intelligence (AI) this year, AI is far from new to financial services. A report published by NVIDIA back in 2022 described AI use as “pervasive,” reporting that across all sectors of financial services—capital markets, investment banking, retail banking and fintech—over 75% of companies were already using one of AI’s core accelerated computing use cases.
While AI may not be new, what is new is the advent of generative AI sophisticated enough for practical application. Generative AI promises to bring game-changing innovations to a wide range of industries.
Before we go further, let’s define the two main types of AI we see in the market today:
-
Traditional (predictive) AI is used to execute rules-based tasks and to identify patterns and trends that can predict future outcomes. Traditional AI has been in use for years—and in some cases for decades—across applications like fraud detection, financial forecasting, healthcare and customer service.
-
Generative AI is a more flexible form of AI that can learn from vast quantities of data, and can then produce original content—whether that’s visual, audio, text or some combination. A conversational chatbot is one example of generative AI.
These two types of AI each have their own strengths. For example, traditional AI is better at complex calculations, so it excels at mathematics-intensive work. On the flipside, generative AI is easy to interact with using everyday language, and its outputs are creative and original. As traditional and generative AI tools each solve different needs, expect them to coexist and complement each other.
Within lending, most applications of AI so far belong to the category of traditional AI. In fact, we can thank predictive AI for some of the most important lending innovations of the past decade, such as advances in underwriting and fraud detection.
Let’s take a look at the example of underwriting. Many fintechs pride themselves on proprietary underwriting models that deliver smarter approvals. While each company may have a unique approach, what most have in common is that they’ve used machine learning algorithms to train their decisioning models on huge proprietary datasets. As more data is gathered, these models improve. The result is that each company possesses a model that, hopefully, performs better than any other for their target market.
To take another example, fraud detection platforms like SentiLink, Unit21, and Sardine are likewise leveraging big data and machine learning to continually train their models to identify patterns of fraud and to refine their approach over time.
While we’ve already benefited from numerous applications of traditional AI, we’ve barely gotten a glimpse of the potential of generative AI. Since there’s no way to perfectly predict all the ways in which AI is going to impact lending, AI will continue to surprise us with its ability to solve important problems—as well as to create new ones. But based on what we know so far, we do have some idea of how AI is on track to reshape the lending landscape. Here are some of the ways in which we expect AI to transform lending.
Fraud detection
Fraud costs lenders billions of dollars each year. And while fraud detection has improved dramatically over the past 10 years, so has fraudsters’ ability to fool lenders. And AI is putting this back-and-forth battle on steroids.
AI-powered fraud detection systems use advanced algorithms to analyze customer behavior, identify suspicious patterns and detect potential fraud in real time. By leveraging machine learning models, lenders can continuously learn from new data and evolve their fraud detection strategies to detect emerging threats.
These AI-powered systems can analyze vast amounts of data, including transaction history, geolocation and behavioral patterns, to identify anomalies and flag potentially fraudulent activity. By proactively detecting and preventing fraud, lenders are doing their best to safeguard their assets, protect their borrowers and build trust. As fraudsters become more sophisticated, lenders will need to make continual investments in the latest technologies to detect, deter and prevent fraud.
Underwriting
Underwriting is another area that’s seen huge advancements in recent years. Historically, lenders relied on manual processes and limited financial data to evaluate creditworthiness. Thanks to the advent of AI, lenders can now analyze vast amounts of data in real time to make better credit decisions. According to Experian, AI has already been able to deliver a 35% decrease in nonperforming loans.
In business lending specifically, AI-powered systems can automate the collection of business financial data and can evaluate complex financial statements, cash flow projections and business performance metrics with greater accuracy and speed. This enables lenders to make faster and better loan decisions, helping businesses access the capital they need and reducing the time and effort required for underwriting.
Unlike fraud detection, underwriting’s adversary is not bad actors but rather the challenge of thoroughly understanding a borrower’s financial situation and accurately predicting how it will evolve. Lenders will never be able to perfectly predict a borrower’s future ability to repay, but expect advances in AI to make lenders better and better at this until there’s little room left for improvement.
While generative AI may help lenders uncover new signals that can help improve underwriting models, it will also need to help lenders with explainability of their models for compliance purposes. With more and more data sources being leveraged for underwriting, regulations will need to evolve in order to help ensure equitable access to credit.
AI will also help lenders to better balance credit risk with borrower needs, enabling lenders to get smarter about offering the right amount of money in the right way. AI may also help increase access for borrowers with limited credit histories, and enable lenders to be more agile in dynamically updating their offers over time. For example, if a borrower’s financial situation improves, a lender could increase the borrower’s credit limit—or upsell or cross-sell—increasing customer lifetime value.
Customer service
Many borrowers prefer to resolve their concerns without interacting with an agent. For all but the most common inquiries, however, the promise of AI-driven customer service has yet to be realized. As often as not, chatbot interactions leave customers feeling frustrated and wishing they had simply been connected to an agent in the first place.
Expect this to change in a big way. Once AI can understand a lender’s system and policies as well as an agent does—and can interact with customers with the same degree of intelligence, empathy and tact possessed by agents—chatbots, queryable help centers and even AI-powered virtual “agents” will finally begin to transform loan servicing. Borrowers will receive quick, personalized solutions and guidance—with no hold music.
Lenders won’t stop there, however. Customer service innovation will continue to advance until customers are receiving much better and more personalized service than they ever have before. Expect lenders to leverage nuanced customer insights and real-time analytics to provide tailored product offerings and personalized guidance. By gaining deeper insight into each customer and leveraging the massive amount of data at their disposal, lenders will be able to make customers feel like they’re using a product designed just for them—increasing customer satisfaction and reducing credit losses. Concurrently, AI must help lenders ensure equal access and fair treatment despite this more personalized approach.
Personal financial management
A wide variety of financial services companies and fintechs have attempted to provide a personal financial management (PFM) software tool to help customers manage their finances. However, none have achieved widespread success due to the difficulty of aggregating separate accounts and providing truly useful, context-aware insights and guidance. AI may be able to solve this challenge. For instance, an AI companion could identify that a borrower has extra cash in their bank account, and could let the borrower know how much they’d be able to save by using that money to pay down their debt faster. Simple wins like this are just the beginning.
Collections and CECL
Expect both traditional and generative AI to supercharge lenders’ ability to collect on delinquent accounts—and to prevent delinquency in the first place. Traditional AI will help lenders both predict delinquencies and optimize their collections cadences to increase recovery rates. Generative AI, on the other hand, will give lenders the ability to generate compelling and personalized communications that will reach borrowers in the right way and at the right time. Expect this AI cocktail to significantly reduce credit losses.
The current expected credit loss (CECL) accounting standard requires many lenders to estimate expected credit losses over the life of loans. AI technologies will play a crucial role in helping lenders minimize credit losses as well as predict them with greater accuracy. Machine learning models can analyze vast amounts of past loan data to identify patterns and predict potential default risks. This will enable lenders to make more accurate financial forecasts, assess credit risk with greater precision, and ensure compliance with CECL requirements. AI-powered tools can also provide real-time monitoring and reporting capabilities, enhancing transparency and regulatory compliance.
Regulatory compliance
The impetus of all lending regulations is hopefully to ensure the greatest good for all. However, after operating in a world of familiar lending constructs like credit cards and personal loans, the arising of constructs like BNPL has raised new questions about how regulations apply. This uncertainty is bad for lenders, borrowers and regulators alike. As technology spurs even greater and more rapid lending innovation, we will almost certainly confront a need to rethink the regulations that govern our financial products, and the speed with which we update these regulations. There will likely be a role for AI to play, both in increasing regulators’ speed in adapting to new innovations and in optimizing regulations themselves given an increasingly complex patchwork of technologies and offerings. AI may also increase the speed with which regulatory bodies like the CFPB can issue guidance to help lenders interpret new regulations.
On the lender side, AI will play a prominent role in helping lenders stay compliant. AI-powered systems can automate compliance monitoring, ensuring that lending practices adhere to regulatory guidelines. And by analyzing loan data, customer interactions and regulatory requirements, AI algorithms can identify potential compliance issues and provide real-time alerts to lenders. This proactive approach will enable lenders to address compliance concerns promptly and mitigate any risks.
A word of caution: Generative AI is prone to the same biases and racism present in the data on which it is trained. So lenders must be mindful of the downstream effects of leveraging AI technologies. Regulators are likely to be wary of the overuse of AI until lenders can demonstrate that their use of AI does not lead to discriminatory downstream effects.
Lending construct innovation
For a long time, we’ve had a number of generally fixed and recognizable product constructs within lending—credit cards, personal loans, merchant cash advances, mortgages, and so on. But these do not necessarily represent the limits of what’s possible. BNPL is a recent case study in how a new lending construct can join innovative technology with novel terms and unique network effects to better meet customer needs. At Peach, we don’t expect BNPL to be the last such novel product construct to arise. On the contrary, we expect AI to accelerate the pace of innovation.
Today, there are a few things that limit lending constructs innovation: regulations, capital and resources, and our own imaginations. We’ve already covered how regulations can evolve to better support innovation. When it comes to capital and resources, AI is already beginning to increase the productivity of many job functions, including software engineering. We can expect that going forward, AI will have a marked impact on the productivity of engineering organizations, enabling lenders and lending platforms to do far more with the resources they have. This will dramatically accelerate the deployment of innovative lending software.
And then when it comes to our own imaginations, it’s not inconceivable that within the next few years AI will be sufficiently intelligent to help us develop even better ideas than we can on our own. This will help lenders better solve the needs of their customers.
In conclusion
While the integration of AI into the lending industry is just one facet of digital transformation, the magnitude of its potential impact—on consumer lending, business lending and commercial lending alike—is hard to overstate. And while no one can predict exactly how AI will reshape lending, we do know that it will give lenders access to capabilities they’ve never had before. Embracing AI will be essential to staying ahead of the curve, driving innovation, and unlocking new opportunities for growth.
Here at Peach, we’re committed to being at the center of this innovation, providing an API-driven ecosystem for our lenders to leverage all the key advances in technology that enable them to gain an edge.
If you’d like to learn more about how we can help you deliver innovative lending programs, contact our sales team.