In recent years, the financial services industry has been innovating technologically, supported by a complex ecosystem including banks, financial service providers, and start-ups (link). Within this blogpost, we showcase our vision of AI in Financial Services.
AI in Financial Services
From our point of view, we can group use cases in AI in three distinct entities, depending on the level of granularity
Microentity Level: At the microentity level, the main purpose of AI is the optimization of operations/transactions without compromising the quality of service. This includes business goals such as minimizing costs and improving user experience.
User level: At user level, the goal is to increase the user value (e.g. by keeping him engaged with the service) and control risk in the services provided to the user.
Company level: At company level, we aim to optimize the company portfolio and return on investments.
What kind of data do I need?
Although most of the use cases described below require specific data sources, we can define a few general data points for each entity. If you work in this area and see a data source you haven’t started acquiring in a structured way yet, get to it!
Microentity: Transactions can be characterized by a value, channel, date, involved parties and other characterization/description (e.g. food, electricity, …)
User: All users should have contractual data (contract start date, contract conditions) as well as behavioral data (customer service tickets, cash flow, …).
Company: A company is characterized by its portfolio size and distribution, external market indicators and economic context, and comparison with competitors.
Companies can also be grouped by its main mission/objective. In each of them, we can further detail use cases for each entity type.
Now, let’s review some specific use cases per company type in the financial services industry. Overall, the use cases focus on using AI for mitigating risks, providing a better experience for the user and guaranteeing business sustainability.
Banking and Money Transfer
Banking and Money Transfer companies transfer money from entity to entity. Companies within this group include Revolut, N26 and Monzo.
Transaction categorization: User transactions can be categorized by business name, location and business type, for instance. Take as an example one product by one of our clients, Pentadata, for merchant identification given a transaction description.
Fraudulent transaction detection: Detecting and blocking fraudulent transactions is critical to increase the confidence of customers in the services.
Predicting recurrent transactions: User transactions can be grouped into recurrent transactions to later on create visualizations on the fixed expenses.
Optimizing spending habits / forecasting months spendings: Based on the historical data of the transactions, it is possible to forecast month spends and make personalized suggestions of spending habits to the user, per category.
Churn Prediction (CRM): Predicting which users are going to churn, and the best action to prevent them from churning, is one of the most common marketing use cases, and can help significantly increase the customer lifetime value.
Upselling (CRM): Determining which services we should upsell the user to, depending on the user behavioral patterns.
Chatbots/Automations: Automating the most frequent questions the users have will lead to reduced spendings in customer service.
Optimize physical store locations:
Forecasting business and market indicators: user growth, net working capital, …
Payment companies transfer money from a person to a company, such as Paypal and Stripe.
Fraudulent transaction detection: detecting and blocking fraudulent transactions – e.g. values out of the ordinary – can help reduce the occurrence of fraudulent transactions.
Predicting card declines: Payments can be declined by the issuing banks for various reasons, such as the card exceeding its credit limit. Predicting and addressing this is one of the use cases Paypal focused on the most using AI (link).
Optimizing user experience: explaining to the users the reason for certain actions in an automated manner (e.g. suspected fraud) leads to increased confidence in the services.
CRM promoting certain behaviors or feature-usage.
Paypal for business / stripe for business fees
Optimizing flows of money: money transfers or currency exchange can be optimized to be performed at the optimal time period, to reduce transaction costs.
Internal investments of funds: deciding on the best way to invest funds.
Compliance detection in reports.
Finance companies – such as Cetelem, Klarna and Cashea, make loans to individuals and businesses.
Claims approval or denial: Automating and detecting errors in claims can reduce manual work and improve the claim processing speed. If you’re curious on how this could be applied in the healthcare domain, take a look at our blogpost we wrote a while back!
Credit scoring: credit scoring uses Artificial Intelligence to predict the likelihood of default based on demographic factors, payment history and other financial indicators. We also have a blogpost on more details on Artificial Intelligence applied to credit scoring. Similar use cases consider the renegotiation of payment conditions and default prediction.
Portfolio optimization: at a company level, companies can aim to best determine the optimal risk level for credits (e.g. spread, short-term, long-term).
If you’re curious about some of these use cases but aren’t sure how beneficial it will be for your company, worry not! In our Data Ignite course, you can find out how to realize potential risks and mitigation strategies at the project conception stage, and learn a common language to discuss AI projects between technical and non-technical teams.
Find out how to realize potential risks and mitigation strategies
Within the investments and brokers groups, we consider companies that facilitate transactions between traders, sellers, or buyers. Examples include DeGiro, trading212, xtb and multiple P2P lending companies (PeerBerry, Mintos, estateguru, GoParity, etc).
Default or delays forecasts: forecasting payment delays/defaults is useful to take action ahead of time.
Trading bots: automating trading decisions at scale
User investment recommendations depending on risk profile of the user, recommend financial products to invest on.
Default prediction: predicting loan default before it happens to better determine the established conditions.
Risk assessment for loan originators: loans typically have a risk level associated, depending with the expected gain/risk trade off. Automating this based on historical data is important to make better decisions.
All CRM-related use cases
Portfolio balancing: determine the best balance between risk and gain for the company’s portfolio.
Lastly, FinTech companies are software companies that provide services to Financial Services, and build part of the use cases listed above. The crypto Industry is also fulfilling the roles of brokers, banking, money transfer, payments, using different technologies.
There are some general use-cases related to KYC (Know your Customer), with general problems such as Legal Document Validation, knowledge tests…
AI in Financial Services is on the path to be a tool to revolutionize the provided services. At the scale of people using financial services, and the fact that most services are online now, facilitating the acquisition of data and the creation of value, there’s a huge potential for innovation and growth in this area.
Special offers, latest news and quality content in your inbox once per month.
Signup single post
Revolutionizing Industry: The Impact of Large Language Models
May 25, 2023 in
Large Language Models (LLMs) are THE hot topic of the year. If the name LLM sounds unfamiliar to you, I’m pretty sure you’ve heard of ChatGPT, OpenAI, and Bard. People who don’t know how to code have gained access to a tool that allows them to build Proof of Concepts for ideas they’ve been meaning […]
The Problem What do you do when the model is underperforming? When the models’ performance does not meet our expectations, we usually spend time searching for the flaws, selecting and analyzing the cases where it failed to understand why it happened. Then, we try to apply more robust solutions, train, test, and repeat. In some […]
The problem So there you are. You have collected your data, analyzed it, processed it, and built your sophisticated model architecture. After many hours of training and evaluating, you have come to a very unpleasant conclusion: you need more data. Before you readjust your budget to fit the extra data acquisition and labeling, let me […]