Unlocking innovation: GenAI across financial services

April 2024  |  FEATURE | BANKING & FINANCE

Financier Worldwide Magazine

April 2024 Issue

For many observers, generative artificial intelligence (GenAI) has the potential to revolutionise the way we live, work, bank and invest, among many other things. Its potential impact has been likened to, and may be as significant as, the advent of the internet or smartphone.

GenAI has exploded, with 2023 arguably its breakout year for grabbing headlines and becoming mainstream. Grand View Research estimated the size of the AI market at $136.6bn in 2022 and expected it to reach $196.6bn in 2023. Further, equity funding in GenAI startups reached $14.1bn in Q2 2023, according to CB Research.

According to Bloomberg Intelligence, growth in the GenAI space could expand at a compound annual growth rate of 42 percent, driven by training infrastructure in the near term and gradually shifting to inference devices for large language models (LLMs), digital ads, specialised software and services in the medium to long term. Rising demand for GenAI products could add about $280bn of new software revenue, driven by specialised assistants, new infrastructure products and copilots that accelerate coding, for example.

According to Google Cloud’s Gen AI Benchmarking Study, 82 percent of organisations considering or currently using GenAI believe it will either significantly change or transform their industry. GenAI has the potential to do so by automating repetitive tasks, enabling rapid prototyping and advancing new product development, as well freeing staff to focus on more profit-generating activities. This will ultimately lead to increased efficiency, reduced costs and accelerated innovation cycles for many businesses and financial institutions (FIs) alike.

Applications and benefits

For financial services including banking and wealth management, GenAI is increasingly popular. These firms recognise GenAI’s immense potential to transform their operations and decision-making processes. A 2023 report by Bain Capital Ventures states that over 75 percent of financial services companies now use AI for their operations. There is burgeoning demand for real-time insights and efficiency boosts.

By leveraging advanced algorithms and deep learning capabilities, firms can analyse vast amounts of financial data, organise unstructured data, identify patterns and generate valuable insights in a fraction of the time it would ordinarily take a human. This not only enhances efficiency but also enables professionals to make more informed decisions based on accurate and up-to-date information. All of this saves time and ultimately leads to better outcomes for clients.

Productivity and operational efficiency are perhaps the most important benefits offered by GenAI. The financial services industry has been utilising GenAI to enhance customer onboarding, strengthen anti-money laundering (AML) and fraud detection, boost marketing activities, tailor money management and personalised finance offers, improve savings and investment advice, assist with credit decisions, and accelerate information extraction and document scanning.

Google Cloud points to five practical use cases for GenAI in the financial services industry, as outlined below.

First, searching and synthesising financial documents. GenAI can help bank employees effectively find and understand information in contracts and other unstructured documents. It can help speed up report generation for bank analysts by researching and summarising thousands of economic data or other statistics. It can also help corporate bankers prepare comprehensive presentation materials for customers.

Second, enhancing virtual assistants. GenAI excels at finding answers within data sets, summarising them, and assisting customer agents or supporting AI chatbots. GenAI-powered chatbots are also more conversational, which improves customer service experiences.

Third, researching capital markets. GenAI tools can serve as research assistants for investment analysts, helping to sift through millions of event transcripts, company filings, macroeconomic reports and regulatory filings, among many others. It can then quickly and intelligently identify and summarise key information contained in these sources.

Fourth, acting as a regulatory code change consultant. With new and changing regulations emerging constantly, financial services firms need to undertake a vast amount of manual, repetitive work to interpret the latest compliance requirements. GenAI can provide developers with context on underlying regulatory or business changes that demand a change of code, show them where to find the answers, and help them to update and check those codes.

Lastly, assisting with personalised financial recommendations. GenAI can help improve customer experience retention and cross sales. It can, for example, create personalised messaging at scale using conversational language and create marketing emails or in-app messages with specific financial recommendations.

Challenges and concerns

Along with all the potential use cases for GenAI in financial services, the highly regulated nature of the industry does pose significant challenges. Trust and accuracy of GenAI are key areas of concern.

According to a recent report by the International Monetary Fund (IMF) on the adoption of GenAI in the financial services industry, the technology shows “great promise” but its intrinsic risks could damage the reputation and soundness of the sector. Those risks are grouped into four key areas: inherent technology risks, performance risks, cyber security threats and financial stability risk. Prudential oversight authorities should therefore increase their monitoring of GenAI development, says the IMF, with interim action needed to help guide use of the technology in the financial services industry.

Data privacy is a primary risk associated with GenAI and large language models (LLMs), given the volume and nature of data required to train the software. Risks may arise, for example, from potential leakage of training datasets or the de-anonymisation of data through deducing identities from behavioural patterns. If users are automatically opted-in to enhance the functionality of GenAI, this increases the risk that sensitive data will inform an LLM and be vulnerable to loss or theft.

Reliability is a serious concern, consistently highlighted as an area that requires considerable advancement in research. For applications such as credit advice and fraud detection, which have zero tolerance for error, FIs must evaluate safeguards and human-in-loop checks to mitigate problems.

Embedded bias in is another consideration. According to Thomson Reuters, GenAI’s propensity toward bias and unfairness is a source of concern: “AI algorithms have shown that they can inherit biases from the data upon which they’re trained, leading to discriminatory outcomes in lending, hiring, and other areas if not properly monitored and addressed. Indeed, the complexity of AI models, particularly generative ones, can make it challenging to understand their decision-making processes, leading to difficulty in explaining the technology’s actions to customers, the market, or regulators. To combat this problem, financial institutions need to actively identify and address biases in the organization’s training data and its algorithms to ensure fair and equitable outcomes.”

If the data used to train LLMs is incomplete or contains underlying societal prejudices, this may perpetuate discrimination in the model’s output. To limit the risk of unethical practices and to maintain public trust in financial services, human oversight is required.

The accuracy of GenAI output is also an issue. Some models have shows their ability to generate novel content which is plausible but incorrect. In some cases they also defend the veracity of this errant content – a phenomenon known as ‘hallucination’. It is not entirely clear why models create such hallucinations, though possible explanations include insufficient data and the complexity of language used. In financial services, the consequences of hallucinations could be disastrous, further eroding trust in these organisations.

Cyber security is also on the radar. A GenAI model may be attacked directly or exploited for malicious purposes. These models may be susceptible to data manipulation attacks, where aspects of data training are altered to potentially weaken accuracy. GenAI can also be leveraged for nefarious phishing communications, to facilitate identity theft or fraud.

The financial services industry is still relatively new to the world of GenAI, so it is difficult for organisations to build a full picture of the cyber threats posed to their models. As GenAI becomes more embedded, and as governments pursue regulatory action, the industry will need to develop protocols and safety principles to ensure data accuracy and security.

Regulatory expectations

Heightened regulation of GenAI is almost certainly on the way. In the UK, for example, there is consensus that not enough is being done in this area. According to research from the Gillmore Centre for Financial Technology at Warwick Business School, 93 percent of senior decision makers at UK FIs believe the UK government should introduce stricter regulations for GenAI. More than three-quarters say they are not happy with the government’s approach to GenAI, and 85 percent have admitted to worrying about the security risks posed by GenAI.

Many of those surveyed have a positive outlook on GenAI’s potential. Ninety-three percent of senior decision makers believe GenAI is going to revolutionise the FinTech and financial services sector, and a further 91 percent are convinced it will play a vital role in financial democratisation. 

Consumers are also demanding greater regulation. While consumers in the UK are divided on the use of GenAI in banking, 62 percent believe that regulation would increase trust, according to FIS. UK survey participants had the least exposure and experience of GenAI of all countries surveyed, including the US, India, Australia and Singapore. Nearly 30 percent said they do not trust GenAI at all, but also confirmed that data transparency, regulation and human oversight would be reassuring factors.

AI is already governed by legislation in most jurisdictions. In the UK, for instance, firms authorised by the Prudential Regulation Authority (PRA) and the Financial Conduct Authority (FCA) are already subject to a wide range of legal requirements and guidance relevant to mitigating risks associated with AI.

The European Union, China and Canada have prescriptive regulations and laws in place. Specific principles apply to financial services firms in countries such as the Netherlands, Hong Kong and Singapore. Cross-sectoral principles, such as the Organisation for Economic Co-operation and Development (OECD) AI principles, promote use of AI that is innovative and trustworthy and that respects human rights and democratic values.

In a more tightly regulated environment, FIs will need to adopt and adapt frameworks that assess and manage risks. A phased approach, based on experimentation and scaling, to support a responsible AI activation framework will be prudent.

Risk and reward

GenAI is tipped to help firms streamline and enhance strategic decision making with added value. In the years to come, GenAI stands ready to unlock innovation and revenue across financial services, driving efficiency, improving customer experiences and increasing regulatory compliance. In 2024 and beyond, financial services firms will continue to explore ways to integrate GenAI technology into their operations.

But its associated risks are still largely obscured, making them difficult to monitor and correct. Meanwhile, authorities will continue to track developments in GenAI and react accordingly.

To deal with the inherent risks, GenAI will need human supervision. Striking the right balance between innovation and risk mitigation will be key to unlocking the full potential of GenAI in financial services. If this is done properly, the rewards could be sizeable.

© Financier Worldwide

BY

Richard Summerfield