Advent of AI & ML Technologies in Today's Finance Landscape

Having completed his MS in Mathematics & Computer Science from Imperial College London, Usman is an experienced finance professional who has handled key roles with UBS, Markit, Goldman Sachs and Lloyds Banking Group, prior to joining Acuity in 2018. In a recent conversation with Siliconindia, Usman Ahmad, Chief Data Scientist, Acuity Knowledge Partners shared his insights on the current fintech ecosystem in India and ways AI and ML are impacting the financial services. Below are a few noteworthy extracts from the exclusive interview

Share your thoughts on the impact of AI and ML on the finance sector lately.

Artificial Intelligence and Machine Learning technologies are reshaping how the finance industry operates by transforming the decision-making process for investment and risk managers. These technologies and quantitative tools sift through and analyze large volumes of data rapidly to allow hidden insights to be uncovered. For example, one of the largest US investment banks has revolutionized its approach to analysing dense legal documents. Leveraging AI and Natural Language Processing (NLP), the bank expedites this process with accuracy and efficiency. A more general example comes to mind which applies to most hedge funds, where the application of AI-powered platforms allows analysts, traders, and strategists to navigate the volatility and complexity of financial markets, attempting to formulate predictive strategies.

What are the major challenges fintech companies face while implementing AI and ML?

The major concern while implementation AI and ML in finance is safeguarding the financial and personal data. With the emergence of Large Language Models (LLMs) which demand vast volumes of data to be effective, gathering data in an age of stringent data privacy laws is going to pose challenges and questions on what type of architecture is best suited to manage LLMs. Another big challenge is the explainability of AI/ML models. The end users of these types of models in the investment sphere are usually bankers, investment analysts and traders who make decisions based on financial reasoning. So, it is very important to provide that financial reasoning when an insight or prediction is called out. Classical ML models such as random forest classifiers or gradient boosting algorithms can have their predictions explained through features importance measurements and local interpretable model-agnostic explanations (LIME), but it becomes more difficult to explain predictions when it comes to artificial neural network algorithms which employ deep learning. Even though deep learning may provide enhanced model performance on unstructured data such as text or images, it is difficult to extract the drivers and interpretation behind the predictions.

Tell us about the various opportunities today’s digital business landscape offer for AI and ML.

Blockchain has emerged as one of the most prominent technology in the current and future digital landscape. In some cases, by integrating AI and ML with blockchain technology, firms are able to devise platforms that can autonomously fine-tune trading strategies via analysis and modelling of vast volumes of real-time market data. The frontier of innovation doesn’t stop at blockchain. The oncoming wave of quantum computing is well suited to add another layer of sophistication to AI and ML in finance. Quantum algorithms could process mountains of data in fractions of a second, offering insights and projections with a degree of accuracy and depth previously considered unattainable. It is an exciting prospect which is currently in a development phase and not ready for wider use mainly because quantum computers currently cannot handle the required number of quantum-bits (q-bits) for AI/ML processing, and there are presently a lot of issues around noise in quantum results leading to errors. These two main blockers are the focus of academia and top financial institutions in this space of quantum computing.

Generally speaking – DARQ technologies (Distributed Ledger Technology, Artificial Intelligence, Extended Reality and Quantum Mechanics) are paving the future road of an exciting digital landscape. When all these technologies come into play, we will see a digital landscape that will have unprecedented AI power with quantum computers coupled with DLT blockchain technology for auditable and secure transaction processing shown and interacted through Augmented (AR) and Virtual Reality (VR).

How has the integration of AI/ML impacted the new economy?

Currently, we are at a stage where the integration of AI/ML in the financial world is accelerating. While it is uncertain if this will lead to a bubble in the new economy, it will no doubt be attractive to the investors. There are many amazing firms out there building AI/ML systems for all sorts of financial use cases – be it providing labelled data, pre-processing services, cloud computing, AI/ML modelling, or reporting UI capabilities. One standout contribution is more from the field of computer science rather than data science with the introduction of cloud computing companies. Synergising AI-driven analytics with cloud infrastructure allows financial institutions to use AI and ML at massive scale with high speeds. This sort of real-time informative technology is naturally going to be appealing to investors and users of this technology which also includes of course financial users.

Please provide your thoughts on how BFSI improves customer experience and retention using AI/ML software.

In my view, there is emphasis on personalization and anticipatory services within the BFSI sector. There are many examples of bank’s developing and using AI and ML frameworks to analyse customer data and transactional data, finding deep insights to anticipate future loan or insurance requirements giving the bank a more proactive stance. This kind of methodology reduces decision-making timeframes for high volumes of customers. Furthermore, this also helps to provide customers a more personalised service helping customers to trust the service providers in understanding their needs.

Parallel to these advancements, the introduction to LLMs in the BFSI world is looking to enhance the nature of customer support. Integrating LLMs into chatbots trained on personal and transactional data can transcend and exceed the limitations of current conventional chatbots by providing that bespoke personalised support and service to a customer.