A&M: AI streamlining compliance in financial services

Fintech Magazine speaks to Rav Hayer, Managing Director at Alvarez & Marsal, on AI’s growing role in augmenting back and front-end operations for financial institutions, as well as how to be wary of using AI in the wrong ways. Read on for Hayer’s take on AI in finance. 

How can AI help financial institutions comply with regulations?

Artificial intelligence will fundamentally change the way the financial services industry is able to comply with the ever-changing regulatory environment it finds itself in.

We’re already seeing how AI integration can ensure financial institutions are complying with existing regulations. For example, a model encompassing an institution's policies, procedures, governance, and regulatory framework can be designed to align with current or new regulations. The model is then able to identify potential gaps in the institution’s approach and recommend necessary amendments, thereby providing a “second line” of defence.

Conversely, can regulations keep up with the technological innovations we’re seeing in the industry? Is there where AI can help?

This is currently unclear. If we look at previously disruptive technologies, like blockchain and digital assets, changes to regulation have been relatively slow. Financial institutions are trying to innovate but regulation sometimes fails to keep pace, due to the need for input from multiple players and the need to align with other frameworks.

Drawing on the lessons learned from previous disruptive tech is important if regulation is to keep up with the innovation that AI is set to unlock. From the outset, the industry will need to establish policies and considerations that prioritise the needs of the customer. We will also need to be quick to adopt and integrate new regulations, even if they’re in draft stages.

And we should let AI helps us on our way here. AI itself can play a supportive role in developing regulatory frameworks, for instance by identifying the processes that are required based on a set of guidelines we provide to it.

How can AI be employed to manage finance functions and help maintain balance sheets?

Firstly, we should be aware that most current AI models struggle with processing tables, graphs, and figures. In most cases, data needs to be transformed into a more comprehensible or natural format for AI to process it.

Having said that, well-trained AI models are efficient at comparing various sources and can identify discrepancies and generate outputs. These models can certainly assist finance functions in auditing financial figures, by detecting inconsistencies in balance sheets or creating tax reports.

However, while we’re in the early stage of these technologies, it is essential to adopt a 'human-in-the-loop' approach where AI serves as a tool, rather than the final decision-maker, particularly in tightly regulated industries like financial services.

Walk us through some of the ways AI can help support AML activities, as well as information verification and underwriting

A well-trained AI model can be used to make decisions that align with previous manual analysis, so could prove invaluable for AML activities, information verification, and underwriting.

In terms of AML, artificial intelligence can be used to detect red flags based on information provided by new customers, such as name similarity or account inconsistencies. In addition to identifying inconsistencies, advanced AI models excel at processing what’s called unstructured data, and so can be used to locate specific information. This makes it extremely useful for information verification activities.

If we turn to underwriting, AI can use historical data to make decisions asked of it, which has the potential to innovate current underwriting methods. For instance, if AI had access to transaction history, it could be used to estimate a client's risk profile and correlate it with their credit score.

How important is it to stay mindful of the downsides of AI, such as its impact on labour forces and IP concerns? How can these downsides be mitigated? 

Very. The effective implementation of AI within financial services remains in its early stages, and the reliability of these models hinges on the quality of their training and development. As we do move forward with AI, we’ll need a rigorous approach that involves all stakeholders (both internal and external) to ensure consistency and crucially to mitigate biases in AI solutions.

The integration of AI technology poses various ethical challenges, for both regulators and financial institutions. Most obviously, AI solutions must clearly indicate their information sources to prevent intellectual property issues, while the encryption of personal data will be essential. Another important consideration is the use of green energy for AI models.

Banks in particular will also need to be upfront about their use of AI and the data used to train their models. Not everyone will embrace the use of AI.

Drawing from historical lessons, the key to navigating the AI revolution lies in promoting sustainable development, responsible regulations, and robust bias control measures.

How can you see AI evolving in the next 5-10 years? 

I see the evolution of AI, specifically, Generative AI (GenAI), transforming both industry and society over the next 5 to 10 years. It promises to have an impact akin to the Industrial Revolution and the launch of the internet.

Several factors will influence GenAI's trajectory in the coming years. The next decade will see AI researchers tapping into diverse data sources, resulting in more realistic AI-generated data and opening up new applications across industries. New data types such as video, voice, and sensory data will play crucial roles in enhancing user experience, crucial for sectors like retail and entertainment.

The ongoing advancement in computational power, coupled with quantum computing developments, will play a key role in overcoming the current limitations of AI. These leaps in processing power will enable complex, large-scale GenAI models to be trained with increased efficiency, unearthing solutions to previously impossible problems.

What must we be mindful of as AI proliferates across multiple sectors?

Establishing global control frameworks and best practices should be top of our priority list as AI proliferates. While we need to promote innovation, more important is the safeguarding of user privacy and preventing misuse.

AI is a global technology, so governments and industries alike will need to collaborate on legislation. The OECD and its privacy guidelines provide an excellent example and potential starting point for defining a broader framework for AI.

We also need to be mindful of GenAI's environmental impact - its energy-intensive nature poses real challenges to our sustainability efforts. That said, AI itself can help us make eco-friendly advancements in energy usage, system efficiency, and renewable energy production, so there is an opportunity here as well as risk.