Fintech is considered one of the fastest growing industries. And it’s not surprising. Anyone who had to deal with a bank in the last couple of years would gladly opt-in for a more customer-friendly financial service, with transparent fees and superior customer service.
It’s much more convenient to transfer money through services like Venmo, then to make sense of the options provided by your local bank. It’s much easier to shop around for a loan, using services like LendingTree, then to actually go to your bank, wait, and then deal with the paperwork.
There are also plenty of options in other fintech segments, from peer-to-peer lending to investment advisory services. With cheaper services, flexibility not available through traditional financial institutions, and incredible convenience, fintech adoption rates are astounding.
At the same time, banks are ramping up investments in fintech-enabled services. They certainly have the capital to make an impact. The internal industry competition also keeps growing, driven by potential opportunities.
This pressure makes fintech companies search for new opportunities to stand out, provide better service, and improve customer experience. That’s why they turn to one of the most abundant resources at their disposal in search of a competitive advantage – data. And there’s plenty of it: browser specs, transaction history, geolocation, unstructured data (images, voice), personal data submitted through the applications, and much more.
But fintech is way past legacy systems, spreadsheets, and pivot tables when it comes to data analysis. So what’s the secret ingredient? Artificial intelligence drives a lot of innovations in fintech analytics and product development. In fact, according to Mediant, most fintech investment professionals consider it to be the most impactful technology in fintech right now.
It is not a magic pill, but people have their reason to consider AI so impactful. It’s likely that AI will enable profit gains well over 20% across multiple industries within the next two decades, with financial services being one of the prime benefactors. With many services being digital-first, advanced data management and analytics are in their pedigree, which pre-determines the potential impact of AI. The challenge is not exclusive to fintech, as other industries are also overwhelmed with data and look for new ways to make sense of it.
Using the power of machine learning or complex AI-based decision systems, AI consultants can help solve most of the pressing data analytics challenges in fintech. That’s why, according to Hays, one of the largest recruiting companies in the UK, AI specialists will be one of the most demanded roles in fintech.
So we wanted to take a closer look at specific applications of AI in fintech and how they can impact business processes and profitability.
Credit Risk Scoring and Underwriting
Most of the fintech companies specialize in services usually reserved for traditional banks. That’s why lending and financial management are the biggest niches within the industry. In the United States alone, consumer credit debt approached 4 trillion dollars. That’s an enormous pie and an excellent opportunity for alternative lenders.
This means that credit risk is one of the hottest venues for AI applications. It’s in high demand. It’s historically based on analyzing consumer data. And it has a direct influence on the profitability of the company. Even more so for companies working exclusively in the lending space.
One of the most significant benefits of improved credit risk assessment is the ability to distribute more loans. And if increased risk appetite is not the final goal for your fintech company, then it merely allows you to improve the quality of your existing loan pipeline.
AI provides a tremendous advantage for fintechs using it for credit risk analytics. Instead of relying on just FICO scores, rule-based loan application reviews, or legacy statistical models, the lender can digest all of the variety of consumer data to build that prediction.
The success of AI in fintech is hard to debate. For example, Lenddo uses not only the loan application data but the whole digital footprint of an applicant to generate a credit risk score. Lenddo’s machine learning combs through social media, browsing, geolocation, and other external data to draw more profound conclusions about credit risk. They claim to improve credit approval rates for some of their clients by up to 50% due to a better understanding of the actual credit risk.
An Australian fintech company Harmoney was able to double the size of its loan offerings and extend the loan periods for its clients by using machine learning to process loan applications. Simply put, they know applicants better because they understand the credit score better. This means that they can trust the borrowers and give them more money for longer periods.
Underwriting can be significantly automated due to such advances in credit risk scoring. Instead of having people manually review applications, the ones that have lower credit risk can be automatically accepted, saving time and resources. According to the Oxford University, 99% of underwriting duties will be automated by 2030. The research talks about insurance underwriting, but given the similarities of the process, this could also be true for fintech. Services like Underwrite.AI are a step in that direction.
For a fintech company, optimizing the price could mean a ton of different things. For example, if you lend, then the price is your interest rate. If you provide transaction services, then the price is the exact fee that you’re looking to charge. These types of use cases for AI can also be referred to as dynamic pricing.
Achieving the price optimum in these kinds of use cases guarantees many advantages:
- You do not overcharge people, thus improving conversion and retention rates. In fact, for banks, 39% of customers are likely to switch to a different provider because of better pricing. The same is true for most lending and mortgage companies in the fintech industry.
- You maximize profits: by not overcharging you reduce the risk of losing prospects.
- You minimize the risk of undercharging, which translates into potential lost opportunities (the amount you might have earned through a higher rate). At the same time, you’re lowering the default risk, as sometimes exorbitant interest rates decrease the chances of the loan to be paid out in full.
Similar to credit risk, AI uses historical data to deliver the results in this use case. However, depending on your specific implementation these problems can be solved through different types of machine learning algorithms (regression, classification, etc.).
There is a variety of ways to tackle pricing. There are publicly available research papers that detail how you can predict the optimal price for your financial services. You can also use established providers in the space, like AWS and their machine learning capabilities, to build price predictions.
Losses from online fraud amount to billions per year. The fintech industry is not an exception, especially for companies that handle transactions and process payments. The biggest problem with legacy systems that detect fraud is that they’re cumbersome, pretty often developed as rule-based systems designed to react to a limited number of potential red flags.
But online fraud is evolving, and it’s always going to be an arms race or a game of catch-up, with fraudsters continuously developing new ways to steal information and money. One of the possible answers to all of these threats is AI. That’s why larger financial companies are already investing in AI. In Q1 of 2018, HSBC invested in Quantexa, which saved the bank more than 7 million pounds within the scope of the pilot project by detecting and preventing fraudulent transactions with machine learning.
There are tons of bootleg solutions if you’re exploring the concept, like GitHub repos with algorithms trained on Kaggle data for credit card fraud. Large enterprises like SAS also offer fraud detection services, but you’d better be ready to shell out some serious cash. However, there are tons of other vendors, with various pricing, services, and compatibility specs.
An AI-based fraud system doesn’t have to rely on rules, which severely limit its purview. It can process all of the incoming information at once and find hidden patterns within data. It’s also called anomaly detection – meaning that machine learning analyzes incoming data (e.g., transactions, payment requests, loan application data, etc.) and spots anomalous behavior.
Cross-Sell and Upsell
Although there are one-hit wonders in the fintech space, offering only one product, most of the providers have a variety of services available. Fintech companies want to sell these services, and the existing customer base is the perfect audience for that. If a person is already using one of your services, it’s likely that they might be interested in other offers.
But what specific customers to target? What products will resonate with them? When marketing and sales teams don’t ask these questions, consequences might be very negative. You risk alienating customers with aggressive marketing and upsell offers. But artificial intelligence is perfect for these types of problems.
AI can be used to maximize the impact of your sales activities by determining clients susceptible to each specific product offer. This also can be referred to as bucketing, clustering, or segmenting. The premise is simple – identify people with similar purchasing habits/ profiles and target each group with tailored product offers. You can manually build these models with tools like scikit-learn. Or find an enterprise-ready solution like RapidMiner.
Selling additional products to your existing customers is also a great way of improving their loyalty. People using more than one product from the same company are less likely to churn. Which brings us to the next point.
Churn rate is probably one of the major KPIs for fintech startups, especially for those in the early stages of their funding. Improving retention by 5% can result in a 25% to 95% improvement in revenue, according to Bain & Company, the inventors of the net promoter score. That’s why improving retention rates should be one of the key strategic goals for any fintech company.
But it’s not just about upselling things to keep people hooked. It’s also about identifying people likely to stop using your services or financial products. There are markers within customer data that can point you in the right direction, like the frequency of transactions or logins. But only AI can identify the full range of these markers and how they’re interconnected.
As with other use cases, this one has a ton of potential vendors or solutions ready to help you with the war against attrition. Tellius offers a comprehensive suite for predictive analytics, powered by AI. Customer churn is only one of the use cases that they support. Microsoft’s Machine Learning Studio offers churn prediction. There are tons of open-source solutions and modeling approaches that can help with this problem, as long as you have the expertise on board to pull it off. If you’re just exploring the topic, then remember that churn/attrition types of problems can also be referred to as survival analysis.
The Tip of the Iceberg
These are just a few use cases for AI in fintech. We decided to focus on the potentially most impactful ones. However, there are many other opportunities and applications of AI in the industry. Let’s briefly go over a couple of examples.
Fintech companies in the investment management space might be interested in analyzing portfolios to provide better financial advice to their customers. Insurtech companies might be interested in identifying policy risks for each new prospect so that they could focus their efforts on the most promising ones. Fintechs heavily invested in marketing can use AI to improve marketing attribution to better understand what marketing efforts are more efficient at converting people.
It’s worth noting that the eventual ROI of AI in your specific case highly depends on your strategic focus and the exact setup that you’ll be using. The quality of the data that you collect is critical, as AI will use that data as its knowledge base. Errors in data or missing data can greatly hinder the learning process for AI.
Not all of the commercially available products out there will work for your fintech company. There are various reasons for this. One of the biggest is the fact that AI is still in its infancy and vendors can’t always tailor their product to individual businesses or business problems. As we mentioned earlier, your readiness for AI is also essential. If your data management practices are lacking, feeding the information to the system might be fruitless or even harmful for your business.
Are you consider an AI project for your fintech company? Do you have a success story for AI in fintech? Share your thoughts and ideas in the comments below!