Managing risk is fundamental to the health of any financial institution, as accurate risk management is essential for driving revenue and fueling new investments. But there are about 1.4 billion credit-invisible consumers in the world who lack sufficient traditional credit data. Through AI and deep learning innovations, fintech startups and financial institutions can now leverage troves of alternative data that go beyond traditional methods of risk management to price risk more quickly and accurately.
Coupling those innovations with the greater visibility of the open banking era, fintechs and financial institutions are able to lessen the cost of banking for previously underserved customers while simultaneously opening up more funds to fuel new investments. With the power of AI and high-performance computing, the financial industry can now take a giant leap towards a more inclusive and equitable financial system.
Accurately assessing risk, faster than ever
Lending and credit has been one of the largest funding areas of fintech over the past 10 years, according to a research by FT Partners. With the use of high-performance computing (HPC) in conjunction with AI, banks can now understand and manage financial risk in near real-time. HPC helps fintechs process large amounts of data quickly, essential for analyzing complex risk factors. This is crucial because in the event of a bank failure, financial institutions need to quickly assess the potential risks to their own balance sheets and customers—especially when billions of dollars worth of deposits are at risk.
Traditional risk scoring places an emphasis on behavioral variables such as historic, late or missed payments. These scores are typically derived from credit bureaus and tend to overlook thin credit files, individuals with no credit history, or those without access to banking services. Modern AI-powered techniques combine historic data behavior and propensity to pay with real-time data to provide a more effective measurement of the real credit risk of a consumer, separate from their payment history.
“Credit in the past was lent to people that had credit in the past. Anyone that doesn’t have previous credit or anyone that’s not part of the financial system wasn’t getting access to credit for those reasons,” explains Sam Edge, Global Head of Fintech for Startups at AWS.
Upstart, an AI-based personal lending marketplace that uses non-traditional variables such as education, employment and social media data to predict creditworthiness, leverages the AWS AI/ML platform powered by NVIDIA’s accelerated computing platform to power their deep learning models for underwriting. In doing so, they’ve been able to reduce model training time by 40%. With deep learning models, Upstart was able to approve 27% more borrowers at a 16% lower average APR.
How AI is reshaping car insurance
The rise of AI underwriting also has big implications for auto insurance. Telematics, for example, is an emerging mechanism for sourcing data that is based on a person’s actual driving habits. Telematics data comes from devices installed in cars, which monitor driving habits and can use GPS to assess how safely a person drives. Insurtechs like Root Insurance have built telematics solutions for mobile devices, making it even easier to implement this cutting-edge technology in any car.
Personalized risk management leads to a healthier balance sheet for both the insurers and the customers. Customers benefit from lower rates because their risk is being priced based on their driving behavior, rather than outdated demographic data and credit scores. Insurers, on the other, benefit from a more accurate understanding of the risk associated with each driver, based on factors such as acceleration, braking, cornering, and time of day driving. This provides better protection for the insurer and more accurate pricing for the customer.
“If you are pricing somebody based on their driving behavior, they’re actually more likely to drive more safely,” says Edge. “You’re incentivizing good behavior because the driver knows that if they’re speeding or if they turn too sharply, that’s going to impact their policy and how much they end up paying for insurance.”
Why alternative data trumps credit reports
Meanwhile, open banking has become a pivotal force behind a new wave of lending innovation, as it allows companies access to data that were previously unavailable. More and more lending platforms seek to streamline the traditional lending process by using AI-enabled models to analyze alternative data sources and rapidly underwrite borrower credit risk, even without historical data available. The use of alternative data allows fintechs to cater to previously underserved segments of the market, and has changed the game when it comes to quantifying risk for people who are unbanked.
An array of non-traditional variables can now complement traditional credit models. Whether a customer uses a Gmail or Hotmail account, their phone charging habits, or their online activity patterns during certain parts of the day, are all factors that have the potential to influence credit models for companies, in the absence of traditional banking data.
Lendio, a loan marketplace that uses NVIDIA and AWS platforms, enables access to capital to underserved communities and small business owners by matching them with a diverse network of lenders, using machine learning and AI to assess risk and optimize pricing. Through the accelerated risk scoring process, small businesses can access capital in as little as 24 hours.
“We firmly believe that lending automation expands approvals and reduces discrimination and bias,” says Abby Sleight, Data Scientist at Lendio. “This is supported by a National Bureau of Economic Research paper that found, when looking at data from Lendio and others from the CARES Act’s Paycheck Protection Program (PPP), the more automated the lender was, the less likely minority business owners were to be discriminated against.”
Across financial services, AI is increasingly essential for success because much of how banks, insurance companies and asset managers differentiate is in the realm of intellectual property. With AI, fintechs can improve on their data and insights to make more informed decisions and identify opportunities that they may have otherwise missed.
“In fintech, it’s not about how fast you can build a widget and get it onto store shelves,” says Kevin Levitt, Head of Global Industry Business Development for the Financial Services industry at Nvidia. “It’s about how many different variables and insights can you leverage in making a decision around acquiring a customer, servicing a customer, and helping a customer along their financial journey.”
Powering ESG goals with AI solutions
Financial inclusion is a core piece of the ESG mission, and fintechs are using AI-powered solutions to extend the reach of the financial system to those who are underbanked or unbanked. With AI-powered underwriting, customers have the ability to obtain a credit card and build their credit without a credit history or annual fees.
At the same time, investors can collect and analyze more information than ever before when accounting for environmental, social, and governance risks and opportunities. Increasingly, it’s also become common for financial institutions to incorporate ESG-related risks into capital reserve calculations—making it all the more important to understand ESG factors quickly and accurately, whether it’s for monitoring satellite imagery to monitor deforestation rates or use natural language processing to analyze company reports for mentions of ESG-related issues.
The sustainability tech platform Clarity AI, for example, uses NVIDIA and AWS platforms to deliver environmental and social insights to investors, consumers, companies and governments. Clarity AI weaves environmental awareness into online shopping by indicating whether a company has lower greenhouse gas emissions than similar businesses, and whether or not a company is transparent in its reporting of climate-related information.
“Our partnership with AWS and NVIDIA has allowed us to perform hundreds of thousands of large-language model inferences per day, providing efficient and accurate predictions for our ESG platform,” says Ron Potok, Head of Data Science at Clarity AI.
Aella, a West African instant loans and microlending company, uses computer vision technology powered by NVIDIA and AWS to instantly verify a loan if the applicant can provide evidence of a reliable income source. The applicant takes a selfie and a picture of a document that confirms their employment and income status, which is then matched using Amazon Rekognition to verify their identity and immediately release the funding.
This is particularly important for African small businesses that rely on microloans to maintain their cash flow. With this technology, they can receive near real-time lending directly to their bank account or mobile phone, allowing them to keep their businesses afloat.
“In an AI-based lending marketplace, you’re able to approve more people at a lower APR, which means it’s more affordable, without increasing risk to the lender.” says Levitt. “That is real financial empowerment.”
CTA: Download the State of AI in Financial Services report to discover key AI trends adopted by financial institutions around the world.