By Nadia Sood1
1CEO and Founder, , CreditEnable
Bill Gates once said that people always overestimate the effect of technology over two years, but completely underestimate its effect over ten. While it is doubtful that he was referring to SME lending, in few sectors could his comments have been more appropriate. Across the world, the introduction of AI into the SME lending space is producing improvements in the most dysfunctional credit market of all: the US$4.5 trillion gap between what creditworthy SMEs need to grow and the finance that banks and other finance providers want to extend to them but cannot, because it has been too difficult for the two parties to transact.
The global market for SME credit stands at a staggering $8.1 trillion.1 In emerging and developed economies alike, roughly two thirds of the working population are employed by SMEs. Providing finance into this sector is critically important for GDP growth, yet historically it has been a slow, costly and difficult process for banks to underwrite these small companies.
That dynamic is now changing because of the introduction of AI. AI is being used across the spectrum of SME lending to solve data asymmetry problems, improve speed and reduce costs in underwriting, improve the robustness of decision-making and reduce risk. The innovation is coming from both FinTechs that are “leapfrogging” banks as well as from companies that are working with banks, and from banks themselves, who are approaching niche segments in entirely new ways with data-driven products.
While banks can draw on decades of historical data to determine to whom they can extend credit, their lending models have traditionally only factored in a narrow range of data points. Machine learning (ML) can help make sense of the vast amount of data a bank holds on its customers. As well as improving the loan approval process, data can flag signs of trouble far more quickly than was previously possible, often identifying shifts in parameters that enable a lender to head off a problem before it happens, ensuring the loan remains safe and the lender remains profitable. Chinese firm JD Finance is helping mainstream banks make use of the data they have to make better decisions. JD uses AI to recognize and analyse 30,000+ risk control variables, 300m+ user credit evaluations, 500+ models and 5000 risk strategies with which to help institutions better analyse risk. Another well-known marketplace for small business loans is Lendio. Its platform surveys 40 unique data points, including time spent in business and monthly revenues, to match the best loan product to an SME applicant. Its successes include boosting acceptance rates by 20% and, crucially, shrinking approval times.
The inordinate waiting period for a loan to be approved is often a make-or-break factor for a small business. After waiting for an initial response, an SME often must carry on waiting for anything from a few days to a few weeks for a decision, and the eventual distribution of the loan. Once again, ML streamlines the process with dramatic effects, creating happier customers, at a lower cost and generating higher profits for lenders. In the UK, Esme Loans (an SME unit of NatWest Bank) has reported that it has hit over £50m of lending to UK small businesses two years after its launch. Its loans over the last 12 months have jumped more than threefold compared to the year before. Esme has partnered with Microsoft to use AI to speed up customer applications and provide AI chatbots to answer common customer questions during the application process, to make faster and better targeted lending decisions and improve customer experience. Alibaba’s FinTech affiliate Ant Financial takes AI in SME lending to new heights. Ant Financial’s entire model depends on AI, using deep learning and advanced algorithms to harvest data. By tracking and analysing spending habits and histories, it determines lending rates and extends credit online through MYbank. Crucially, it also shares its tech with traditional financial firms to improve their own credit offer, boosting loan-making and risk-taking efficiencies across the whole financial services industry.
Not only are no two companies the same, neither are the sectors in which they operate. Location, industry, market forces and demographics are examples of the variables that ML can integrate to improve the outcomes of loan applications and proactively identify new opportunities. By monitoring customer behaviours within sectors, ML can predict the best time to approach a specific SME in a given sector with a specified amount of credit. Certain companies, such as CreditEnable, build integrated AI solutions which combine automatic data capture, data mining, algorithms, rule-based and statistical anomaly detection, random forest model ML, and Bayesian and neural-network-based natural language processing (NPL), to objectively categorize the creditworthiness and trustworthiness of SMEs that that do not have credit scores or ratings. These companies can then help lenders identify SMEs that match their lending criteria, have a high appetite for debt, can repay the debt and are trustworthy in a matter of seconds, thereby reducing the time it takes to pre-qualify an SME from 3–4 weeks and significantly improving the quality of lender loan books.
In the fullness of time, AI will not only be able to help improve the speed of loan approvals but also reduce defaults because it can instantaneously correlate market events that may cause problems to borrowers caught up in economic cycles. AI lenders and borrowers both stand to gain from the added insight and foresight that is key to avoiding situations that lead to defaults.
SME lending is a tailor-made example of where AI can be applied, adding incredible value for borrower and lender alike, just as it has for doctor and patient, traveller and destination. Success in SME lending depends essentially on the lender’s ability to make informed decisions about a company’s future financial performance. What better realm could there be to introduce AI than here were the integration of large data sets not only improves decision-making but also helps create opportunities for lender and borrower. Even social media and utility information is being captured by AI systems that can contextualize loans against market conditions and a lender’s bespoke risk parameters. For example, MyBucks is a Luxembourg-based FinTech that provides loans in seven African countries as well as in Poland and Spain. The company compares the applicant’s social media feed against the information in their mobile wallet. They look at behavioural traits in order to make sure the information on the customer’s cell phone and their social media accounts tie together coherently.
Today, the concern for established lenders should be how quickly niche FinTech providers will be able to use their technological capability to expand both horizontally and vertically.
Established lenders who shy away from AI would be well advised to remain mindful of the trajectory of the formerly humble online bookseller that is now the world’s largest company.
The promise of AI in SME lending is not merely one of technological advances, but of shifting assumptions and convictions: a shift that is allowing rapid and robust decision-making to occur in the service of delivering credit as a way of fundamentally improving the health of entire economies for the benefit of all.