AI in Finserv: Predictive Analytics to Inclusive Banking

Artificial intelligence is revolutionising the financial services industry, ushering in a new era of data-driven decision-making and customer-centric solutions. From predicting market trends to detecting fraudulent transactions in real-time, AI algorithms are enhancing operational efficiency and risk management across the sector. 

Chatbots powered by advanced language models are redefining digital banking experiences, offering personalised assistance round the clock. Perhaps most significantly, AI-driven credit scoring systems are broadening financial inclusion by considering a wider range of factors beyond traditional credit histories. 

In this roundtable, we talk to industry experts on the key areas where AI is evolving the financial services space. 

Meet our speakers: 

• Prashant Jajodia, Managing Partner, Financial Services Sector Leader at IBM UK&I Consulting
• Jamil Jiva, Global Head of Asset Management, Linedata
• Marco Santos, co-CEO at GFT 
• Dr Scott Zoldi, Chief Analytics Officer at FICO
• Peter Pugh Jones, Director, Financial Services at Confluent

AI is being leveraged for predictive analytics to forecast market trends, consumer behaviour and economic conditions. How is it being applied to deliver these predictive insights?

Prashant Jajodia: AI is being utilised for predictive analytics by employing advanced algorithms and machine learning techniques to analyse historical and current data related to financial services trends, consumer behaviours and economic conditions. 

These models learn patterns and relationships within the data, enabling them to predict future outcomes with increasing accuracy. By integrating AI into financial services, businesses can make informed decisions based on data-driven insights, ultimately optimising their strategies and enhancing operational efficiencies.

Jamil Jiva: AI delivers predictive insights through machine learning algorithms that process large datasets, both internal and external, finding patterns and trends that would otherwise remain hidden. 

In the financial sector, firms are applying AI across the board, from compliant investment decision-making to portfolio management. For example, at Linedata we have developed several solutions to help financial advisors and lending professionals to find the best offer or action for their clients. 

AI identifies personalised suggestions of content, actions (i.e. which customer to call) products or investment portfolios which are made available through CRMs, portfolio management systems and financing software.

Marco Santos: AI’s ability to synthesise vast amounts of data allows organisations to connect data from previously disparate sources, and then analyse it to detect historical patterns and deliver forward-looking insights. 

In the banking industry, this is happening at both a high level through traditional data analysis, and, increasingly, through more advanced AI tools including Natural Language Processing (NLP) and Machine Learning (ML). As organisations continue gathering these predictive analytics, many are also in the process of providing feedback to their AI systems which will ultimately improve their predictive accuracy over time. 

The main use case in which banks are currently seeing the biggest impact from AI-powered predictive insights is in forecasting consumer behaviour. For example, using AI to analyse decades of historical data around consumers’ banking habits in the context of certain economic climates, banks can use the resulting insights to decide on new products or services to roll out in advance of shifting customer demands. 

AI algorithms are being used to develop sophisticated anti-fraud systems that can analyse transaction patterns in real-time to detect anomalies and prevent financial losses. How do these algorithms work?

Prashant Jajodia: These algorithms typically employ machine learning techniques, specifically supervised learning methods, to analyse historical transactions and identify patterns associated with fraudulent activities. 

They learn from labelled datasets containing both legitimate and fraudulent transactions, enabling them to classify subsequent transactions based on their similarity to these learned patterns. As new data becomes available, the algorithms retrain themselves to adapt to evolving fraud patterns.

Jamil Jiva: The most efficient AI models leverage reconciliated 360° views of clients based on a company’s front, middle and back-office information. It collects KYC information, historical operations and transactions, and external news and sanction data to understand the transaction profile of each client and identify diverging patterns. 

Learning algorithms are fed with examples of these client profiles as well as a series of transactions which for some profiles are tagged as fraudulent (generally captured by human and rules systems). Modern algorithms then use this data to learn to identify fraudulent transactions. 

Through supervised learning (identifying transactions that resemble previously labelled transactions), anomaly detection (identifying transactions that diverge from normal transactions, for example a larger amount to an unusual destination) and sequence modelling (predicting the next suggested transactions and identifying the real operations that diverge from the predicted ones) banks or firms like Linedata are better able to spot financial fraud and money laundering.

Marco Santos: Imagine a digital detective, always on the lookout for suspicious activity. In a nutshell, this is how AI algorithms are being applied to detect and prevent financial fraud. 

The underlying basis of these AI algorithms is a significant amount of historical data, including information about what constitutes a fraudulent transaction versus a normal transaction. From there, organisations can use both supervised and autonomous training techniques to guide the algorithms to discern between anomalous and non-anomalous transactions. 

For example, if the algorithm automatically flags a legitimate transaction as fraudulent, AI teams can step in with data annotation techniques that redirect the algorithm and prevent the same mistake from happening again in a real-world scenario. 

AI-powered fraud detection algorithms can analyse vast amounts of transaction data in real-time at a scale that’s unattainable by humans. The real-time nature of these systems also allows organisations to prevent loss by intercepting anomalous transactions before they’re settled. 

This scalable, automatic approach also makes it easier for financial organisations to stay in compliance with relevant anti-money laundering (AML) and anti-terrorist financing regulations and avoid steep penalties. 

Dr Scott Zoldi: Real-time anti-fraud systems are developed around the concept of behaviour analytics. These specialised analytics focus on defining real-time recursively updated anti-fraud features, where the state of the feature is only updated based on the previous estimate and current transaction.  

Commonly known as real-time transaction profiles, these profiles contain sets of recursive variables that are retrieved when a new transaction is received, updated and then a neural network combines the various profile variables into a score used to determine whether to block the transaction for fraud.  

The types of real-time profile variables include usage anomaly variables such as acceleration in spend or frequency of transactions, collaborative profiles which quantify the likelihood of a future transaction and behaviour sorted lists which track favourite repeated recurrences in the past history such as favorite merchants, ATMs, destination account and the like to establish what normal for the individual to reduce false positives.  All this is done in real-time to establish safety in real-time payment systems.  

How have AI chatbots revolutionised the consumer digital banking experience?

Prashant Jajodia: AI chatbots have transformed the consumer digital banking experience by enabling 24/7 availability, multilingual support and instant problem-solving abilities. They can understand and respond to queries in real time, freeing up human resources for more complex tasks. 

Examples include Natwest’s implementation of a watsonx assistant to provide digital mortgage support, which has resulted in the business seeing a 20% improvement in customer loyalty and call duration dropping by 10% with staff now having access to more support. As well as Virgin Money’s chatbot Redi which was built by IBM Consulting on Microsoft OpenAI and has a higher NPS score than some of the traditional channels.

Jamil Jiva: As clients and decision-makers come to expect more from chatbots, it becomes more important to ensure that AI chatbots provide high-quality answers from existing internal knowledge bases (using a technique known as retrieval augmented generation). 

Chatbots are also now expected to help financial services institutions to keep up-to-date comprehensive customer profiles by automating the collection of customer information from account and conversation history (such as KYC information, future plans and life moments). 

This makes it possible to automate more complex tasks and suggest products or content. This works particularly well in a hybrid model where chatbots are on hand to help human agents be more productive, and provide personalised recommendations and financial advice.

Peter Pugh Jones: AI chatbots have transformed the consumer digital banking experience, making banking more seamless, personalised and secure. These chatbots, powered by AI and machine learning, improve the user experience by providing real-time assistance, automating routine tasks and offering tailored financial advice.

Unlike traditional banking, where interactions could often be clunky and time-consuming, AI chatbots allow customers to manage their finances efficiently from their mobile devices. For instance, chatbots can automatically categorise transactions, suggest personalised savings goals and even remind users to pay bills before they are due, without compromising on all security.

What’s more – chatbots can offer personalised financial guidance based on individual spending habits, ensuring that the banking experience is not only convenient but also highly relevant to each user’s needs.

All of this creates a much smoother, more efficient and more intuitive experience for consumers. This is particularly evident in the rise of digital-only banks, which leverage AI to provide innovative services that were traditionally limited to in-person interactions.

The result is a banking experience that is just as much about empowering customers with the tools and insights they need to make informed financial decisions, all while maintaining a high level of security and trust, as it is about completing transactions.

However, while AI chatbots have enhanced the digital banking experience, they are not a complete replacement for in-person services, which remain essential for complex financial decisions and for customers who prefer or require face-to-face interactions.

Marco Santos: We’re only just beginning to scratch the surface of how AI chatbots can transform digital banking. 

Generative AI has delivered significant improvements in this area, making it possible for organisations to overcome previous challenges with AI chatbots such as scalability, and making them feel as natural as speaking with a human. 

However, banks and financial institutions are now facing a new challenge in the form of AI ‘hallucinations’, which deliver false information to end users when AI models haven’t been trained properly or have bias built in. 

Especially in highly-regulated industries such as finance where consumers’ hard-earned money is at stake, many organisations are taking their time to ensure that this issue is addressed first, before introducing any further innovations in their AI chatbot experiences.

AI-powered credit scoring systems are revolutionising lending practices by providing more accurate risk assessments. How has this helped with financial inclusion efforts?

Prashant Jajodia: AI-powered credit scoring systems are revolutionising lending practices by providing more accurate risk assessments. How has this helped with financial inclusion efforts? 

Generative AI models trained on large datasets can analyse various factors influencing creditworthiness, resulting in fairer and more inclusive credit evaluations. These models can identify patterns and relationships that might be overlooked by traditional credit scoring methods, ultimately expanding financial access for underserved populations.

Jamil Jiva: AI-powered credit scoring systems are eroding barriers to financial inclusion by assessing a broader range of data beyond traditional credit histories. By analysing transactional patterns and alternative financial behaviours, these systems allow lenders to evaluate the creditworthiness of individuals who may have been excluded under traditional models, such as those without formal credit histories or from underserved communities. 

This approach will open up access to loans and financial services for populations previously overlooked.

As AI technology becomes more refined, lenders can make smarter, more informed decisions in real-time, facilitating equitable access to credit. AI won’t only improve operational efficiency; it will extend financial inclusion across the board.

To read the full story in the magazine click HERE

Make sure you check out the latest edition of FinTech Magazine and also sign up to our global conference series - FinTech LIVE 2024

FinTech Magazine is a BizClik brand ​​​​​​​