About ten years ago, Marc Andreessen famously stated that software is eating the world. He was right. Software is a valuable component of our everyday, modern lives. Fast forward to 2020 and we are witnesses to the fact that AI is eating software. AI is used for personal virtual assistants, intelligent search engines, recommendation engines, self-driving cars, etc.  What happened in the last decade? Well, the field of Artificial Intelligence faced unprecedented growth. An increase in the available computing power and training data combined with research advancements accelerated potential AI use. Motivated by an increase in efficiency, a lot of businesses use AI to automate and optimize their processes. AI impacted education, retail, healthcare, pharma, and transportation industries. FinTech is not an exception. The total economic impact of AI in the period to 2030 is estimated to $15.7 trillion according to the PwC report.

One of the key findings of the survey conducted by WEF is that AI is expected to become an essential business driver across the Financial Services industry in the short run, where 77% of all respondents think that AI will have high or very high value to their business within two years. Honestly, 2030 is quite far away, especially in the information age where everything changes so fast. But recall an old Chinese proverb: The best time to plant a tree was 20 years ago. The second best time is now. With this proverb in mind, let's see how FinTech companies can embrace AI technologies to improve their core business and recognize AI-driven business cases with attractive ROI.

Are You AI-ready?

Before we dive deep into compelling AI business cases let's discuss some requirements. As you may already know, in the AI world, data is the king and the main component for making intelligent decisions. Some important data-related actions must be taken for the effective use of AI.  Monica Rogati's AI Hierarchy of Needs illustrates this idea nicely.

The AI Hierarchy of Needs Pyramid. Source: “THE AI HIERARCHY OF NEEDS” MONICA ROGATI.

From a very high-level perspective, we can recognize three levels of data readiness: 1) no data at all; 2) unorganized, messy data; 3) clean, prepared data, ready for further use. If you do not have data at all it makes sense to first think about data strategy. A sound data strategy should address questions like what to collect, or how to clean and organize data. If you have some data collected, but it is messy, unorganized, or hardly-accessible, it makes sense to consider data lake solutions before starting the AI use cases. Finally, if you have clean, organized, easily-accessible data, you have satisfied the main requirement for AI projects. Before jump-starting complex AI business cases, make sure that you are AI-ready.

How is AI used in FinTech?

In most industries, including financial services, general decision-making is traditionally achieved by human intelligence. Sometimes the decision-making process is simple and revolves around ordinary tasks. Sometimes deciding on something involves intense thinking and processing of complex information. Unfortunately, a lot of cognitive biases affect our decision-making. Humans make different decisions in the same situations, depending on the current mood or personal experiences. Also, humans struggle with complicated patterns and pieces of information because our working memory is limited. Fortunately, some of these flaws are not present in computers and AI. Machines are not only less error-prone than humans in some tasks, they also operate at a lower cost and scale more efficiently. Still, humans are an essential part of the equation. For example, when regulatory risks are present, humans should be part of the loop and AI algorithms should augment the decision-making process. Next, we show how AI can be used in FinTech to personalize the customer experience and automate inefficient time-consuming processes. High-quality data combined with modern AI algorithms can hugely improve financial services.

Competitive Analysis identifies and evaluates the business strategies of your competitors, resulting in the analysis of strengths, weaknesses, opportunities and threats (SWOT) for your product relative to the competitors’ in a business ecosystem. Further analysis may provide an insight to your product strategy.

The analysis is often conducted in the early stages of product development. As the dynamics of products in the ecosystem change rapidly, many companies have embraced agile competitive analysis as a part of their product strategy.
Photo by UX Indonesia / Unsplash

Customer Acquisition and Retention

Acquiring a new customer is expensive. Businesses are exploring ways to reliably enhance customer acquisition. One way is to examine the existing customer base and identify potentially profitable segments. Segments can serve as a starting point for the marketing and sales campaigns. For example, a sales team can precisely target the customer segment most likely to upgrade their subscription plan.

Most probably, your business offers multiple services to customers. If done right, you can cross-sell a customer into higher-margin products. On the other hand, the wrong offers will not result in happy customers. By mapping the customer journey, you can use data to offer a logical sequence of products to your customers. But how does AI come into play here? Well, creating fine-grained customer segments based on complex data is a perfect use case for AI algorithms. Discovering these segments by hand is exceptionally tiresome. In some scenarios, it is impossible to perform a detailed analysis manually. Deciding if some customers should be offered a product or not also involves a thorough analysis of heterogeneous data, a task ideal for AI algorithms.

Imagine if some tool can tell you in advance how likely your customer is to leave your platform. What would you do if a high-value customer is in question? Probably, you will take any necessary actions to keep a customer. There are numerous AI algorithms used for developing such tools. The described problem is also known as customer churn prediction. More precisely, we are addressing the following question: Given the customer's historical data, what is the probability of customer discontinuing service? The customer churn prediction tool can help us preserve customers by taking appropriate actions promptly. As you may already know, keeping existing customers is orders of magnitude cheaper than acquiring a new one.

Automated Customer Support

Another fascinating area where AI shines is automated customer support and chatbots. There are tons of chatbots, designed to be relatable and have a personal touch, enhancing customer support processes. The use of Natural Language Processing (NLP), the subset of AI aimed at recognizing human language and speech, has advanced the development of chatbots to the level where they can perform an impressive set of operations. Automated support case processing or virtual assistants are some examples where NLP techniques turned the daunting experience into a pleasant one. Tech-savvy generation prefers digital services as opposed to standing in long lines and unpleasant interaction with bank staff. Some studies showed that 92% of US millennials prefer a bank with digital services over a bank that does not offer digital options. It is clear how AI-powered services bring value here. The operation cost of chatbots and automated customer support systems is much lower than having a customer support team. Besides, these systems are available 24/7 to customers. On top of that, a seamless customer support experience will lead to more satisfied customers.

vintage telephone on the wall.
Photo by Pavan Trikutam / Unsplash

KYC Automation

The standard KYC process flow includes sequential steps taken by individuals or teams. That is time-consuming and slows the client onboarding down. A lot of manual work has to be done, which is costly and error-prone: manual data entry, verification, document quality check, initial risk estimation, etc. An inefficient KYC process exposes organizations to financial, reputational, and competitive risks. This process can take days or even weeks. Fortunately, AI advancements enable changes that offer a tremendous advantage at a much lower cost than traditional methods.  If you break down the KYC process to steps, you can identify areas best handled by people, and areas that can be automated. The most efficient process also involves highly skilled professionals to handle tasks that require human intelligence and rationalizing. For example, AI automation can do heavy-lifting, giving more time to high-value decision-makers, analysts, etc. Also, AI methods can help with risk prioritization to concentrate limited resources where they are most needed. Which KYC process steps can be automated? Well, almost any tedious, manual work that does not require meaningful critical thinking and reasoning.

For example, the onboarding process is prone to automation because it includes a lot of manual data collection and verification. Steps that include document parsing, named entity recognition (for example client name, address, etc.), date extraction, or entity matching can be automated with current NLP techniques. Most KYC processes have identity verification by government-issued documents like passport or identification number. Often these documents are uploaded as images to digital services and examined manually. In this case, AI technologies like Computer Vision and Optical Character Recognition can speed up the verification process drastically. Manual fraud detection is not very reliable because digital documents are easily manipulated to deceive humans and can cause fraudulent transactions. AI-augmented fraud detection can significantly reduce the possibility of fraud by helping humans detect suspicious patterns during client onboarding. Also, initial client risk assessment can utilize existing AI technologies to analyze data provided during onboarding and classify clients into predefined risk groups. Such analysis can pinpoint potential high-risk clients early and provide additional information for human experts. KYC best practices suggest continuous monitoring of existing customers, and as the number of new customers grows, manual KYC processes would quickly become too expensive or impossible to scale. A successful KYC process needs AI-powered automation.

Personalized Financial Advisors

Also known as robo-advisors, AI financial advisors are one of the highest potential business cases. The robo-advisor objective is simple: dynamically manage available finances and assets to optimize for the client's goals (for example, saving for retirement). Intelligent advisors can guide customers on how to save, how to manage taxes, when to pay off debt, and how to invest. Such a service could drastically enhance the customer experience on the platform. Customers will stay loyal to service not because it is hard to walk away, but because benefits are better than anywhere else. Analysis of customer personal information, spending and saving patterns, markets, risk exposure, tax laws, inflation, etc. provides the best advice possible. The number of relevant data points quickly becomes too large for manual examination. Once again, AI shines in complex pattern analysis. But significant challenges are earning customer trust and regulatory approval. Therefore, robo-advisors could augment human asset managers with more precise insights from the study of heterogeneous data. One may think about data privacy concerns. Surprisingly (or not), Accenture found that 80% of consumers are willing to share their private data for more benefits. These benefits include lower prices, more relevant, personalized offers, and alerts. Data and AI technologies are ready, but customer acceptance would need to increase.

Photo by Rock'n Roll Monkey / Unsplash

Credit Risk Scoring

The potential risk of lending money, or providing credit to someone, is evaluated with credit scores. The traditional credit scoring methods weigh many factors, such as a client's payment history, personal information, income, debt, or even previous credit history, if available. Usually, manual scoring takes lots of time and often separates clients in broad segments without considering individual cases thoroughly. Available credit score data and existing AI algorithms enable a more granular and detailed evaluation of each borrower. With a larger volume of data, AI-based credit scoring can grant credit to clients who would not otherwise have access to it. In cases when credit-related data is not available, AI methods like NLP can analyze social media data for additional insights. Similar to the previous use cases, the benefits of automation are cost savings and easier scaling. Also, if the regulatory burden is present, automated credit risk scoring should aid human decision-makers.

What’s Next?

The impact of AI on the global economy is unquestionable. The AI investments are increasing worldwide, while some experts think it will increase even more. In the FinTech industry, the key areas where AI can cause the highest impact are personalization and automation of decision-making processes. But there are still some challenges slowing the adoption rate like customer acceptance or regulatory compliance. Another set of challenges include attracting multidisciplinary teams with deep experience in data-intensive systems, AI, and financial services. The companies most likely to succeed in AI application to financial services are the ones with AI-friendly executives and experienced AI teams with domain knowledge, capable of driving AI-powered FinTech solutions.