Fintech companies from digital banks to payment gateway providers and stock-trading apps are increasingly harnessing the power of Artificial Intelligence (AI) to automate workflows, improve decision making and add value.
The advantages of AI for fintech are that it can handle data and create data models more effectively than humans, but the disadvantages include a lack of depth of understanding, questionable diversity inclusion and problems with financial regulation.
The June 2020 edition of Research and Market’s report “AI in Fintech Market - Growth, Trends, Forecasts (2020 - 2025)” stated that the AI Fintech market was worth $6.67 billion in 2019, and that it was expected to grow to $22.6 billion by 2025. That’s a compound annual growth rate (CAGR) of 23.7% between 2020 and 2025.
The value of AI in the finance sector is set to increase so dramatically over the next 5 years because Artificial Intelligence is recognized as being more powerful than the traditional data analytics capabilities of financial institutions. This is because:
AI can use raw data. AI is able to take unstructured data, uncover the salient datapoints in it and turn it into a form that it is able to process effectively. One example of this the way that the AI tech Optical Character Recognition (OCR) is able to scan images of documents like contracts or ID cards and extract words and numbers from it to transform into a dataset.
Data storage is flexible for AI. The data doesn’t necessarily need to be cleaned and stored in an orderly way as it is for other types of computer programs, like in a Data Management Platform (DMP). The AI-based data analyst can recognize patterns from raw data that is streamed directly through its neural network.
This means you don’t have to store data in a structured way like in a Data Warehouse, which takes time. Instead you can just dump it in a Data Lake and let the AI analysis tool do the hard work further on in the funnel.
See: Data Warehouses vs. Data Lakes
The data doesn’t have to be perfect. While some software models will deliver incomplete insights if the dataset is incomplete, AI models are able to use Machine Learning (ML) to extrapolate the missing information from the various sources it is fed.
One method of data linkage where there is a gap in the dataset is through Probabilistic Matching, which compares the statistical similarity of datapoints, for example to identify one user across multiple channels or contact points. With ML, the computer can learn from its mistakes to improve the accuracy of probabilistic matching.
AI can learn from its mistakes without humans. One vital feature of AI is that it can detect changes in data patterns and tune the analytics model accordingly all by itself, without human intervention. You don’t need data scientists to come in and tweak the model after an anomaly has been detected. Automating the data modelling process in this way accelerates business workflows and ultimately saves the company money.
AI is reaching further into our lives… will you take its hand or give it the finger?
There are two sides to every story, and Artificial Intelligence in financial technology is no exception. These are the disadvantages of AI use in fintech:
AI is exclusive. One example of the way AI systems are vulnerable to biases is in their treatment of disabled people. AI and other new tech solutions are often built to serve the needs of the many and the privileged, and not with inclusivity and diversity in mind. AI is not woke. For instance, take eKYC or digital Know Your Customer, the process of using image scanning or voice recognition software to confirm users’ identities. As technologies of this kind become better attuned and more accurate, they are increasingly made into obligatory steps in the financial security measures.
But for blind and/or mute people, or those with limited mobility who find it difficult to operate mobile phones and traditional technology, these AI projects fail to account for their differences and cannot recognize them, excluding them from AI-powered services. Facial recognition software is also notoriously racist, having picked up on biased attitudes in our society to profile certain races as criminal or untrustworthy. These are just two simple cases in which AI is dangerous because it carries an in-built tendency to exclude minorities.
Shallow understanding is more common than deep learning. The most basic form of learning in humans and animals is imitation, and Artificial Intelligence is very good at this. But imitating human thought can be done by humans without the need for expensive machines. Part of the reason we look to develop AI is to be better and smarter than humans, not merely to copy us and do exactly what we do but faster.
The challenge is teaching AI to learn for itself, to develop a deep understanding of problems and innovative solutions, instead of just a shallow, mechanical intelligence. Developing these systems of thought in machines will mean thinking outside the box for their human creators, to program a very different kind of learning and thought process than we’re used to, and this capability is still not available for AI.
AI in finance is unregulated. The finance and banking sectors are regulated by strict laws that ensure fair play and compliance with tax codes. However, as Paul Sin, Deloitte AP Blockchain Lab Leader points out, AI analytics for financial services is different from traditional, human-powered analytics because it cannot be used for regulatory reporting. This reason is that it is not possible to extract the propensity model from the neural network to show to the regulator. This barrier to legal regulation of fintech when AI is used may lend itself to law-breaking and corruption by the financial institution, either intentionally or unintentionally.
The pros and cons of AI in fintech solutions rely heavily on the capabilities and limitations of the technology. But AI is improving all the time and as the machines become more powerful, these disadvantages will be mitigated or extinguished. That’s not to say that new problems won’t arise, but progress demands the development of creative solutions to these problems.
It’s inevitable that AI will become more and more prevalent in the financial sector, being used on a daily basis to aid productivity and help in decision making. Data scientists, fintech providers and robotics engineers need to work together to ensure the conscious application of technology to overcome obstacles and make financial services work for the good of the planet and all people on it equitably.