Keeping pace with IT is a challenge in any industry, however financial services comes with a unique set of challenges, objectives and strict compliance requirements. Staying relevant is essential for maximising gains. AI has been maturing and has evolved rapidly to the point where it has become accessible to businesses of all shapes and sizes. The emergence of out-of-the-box solutions has empowered smaller firms to embrace this technology trend in unprecedented ways.
An explosion of use cases
Today in Financial Services, AI is mainly used for risk and pricing, however, something I have already written about, in the form of generative AI, holds immense potential to broaden use cases across a wide array of new areas:
- AI can streamline software and programming tasks through automation and optimisation.
- AI-powered chatbots enhance customer service with a personalised experience and prompt problem resolution.
- AI algorithms can assess credit risk and provide a greater level of accuracy to the underwriting process.
- Anomaly detection in real time mitigates risk and minimises fraud.
- AI can identify investment potential and personalise portfolios.
- With fast and powerful analysis, AI can summarise vast amounts of information.
- Search and retrieval can be automated.
- AI bots may well carry out financial activities and transactions in the future.
As AI has become more sophisticated, the application of models in finance has broadened, with these use cases becoming more common.
- Combatting financial crime
AI-powered scams include chatbots that mimic human interactions and deepfakes. In Q1 of 2024, the average weekly cyber-attack against financial organisation reached 1,172. In the modern age you must fight fire with fire by using AI! AI based real-time analysis of vast datasets means accurately identifying suspicious activity and detecting fraud is not only possible, but possible at extreme scale. For example, Mastercard’s tool can scan up to 1 trillion data points and deliver a safety score in under 50 milliseconds, potentially reducing false positives by up to 85%.
- Streamlining coding tasks
As someone with a coding background, I can see both huge benefit and potential risks when it comes to code generation. Used with developer culture in mind, there is great potential for AI to help with productivity throughout the software development process. From creating and optimising code, detecting errors and automating tasks, a study showed a productivity increase of 56%, when programmers were using AI tools. However, get it wrong and you run the risk of developer revolt - no one likes refactoring AI generated code! Also, as AI generated code becomes more complex, there can be challenges during downstream activities, such as testing, where understanding the code base means comprehensive test coverage.
- Making better decisions
Traditional data and statistical models have historically been used to calculate risk, however AI could increase the accuracy of these predictions by analysing larger datasets. AI can look back through vast amounts of data from a range of sources, minimising manual processes and reducing errors. This leads to more precise risk evaluations and faster policy underwriting. A leading insurance company has lowered premiums by 5-10% thanks to a more accurate understanding of risk. Remember to avoid the challenges posed by historical datasets that do not have the right class balances, missing or anomalous values before hitting the scale button.
Ensuring compliance
AI regulation is still evolving but is gaining traction on global agendas. The first ever AI legislation, the EU AI Act (check out my primer if you are unfamiliar with the act), has lead the way in AI regulation with the G7 following fast. This makes things even more complex for the financial services sector who already face stringent compliance requirements and regulatory frameworks.
If an AI system isn’t auditable then proving compliance is impossible. High-risk AI use cases in regulated industries are already restricted, but for lower risk use cases, steps can be taken to make AI auditable and therefore compliant.
Creating an auditable AI system requires clear evidence of lineage and traceability. This means documenting and storing meta data about changes, who made them, why they made them and ensuring clear version control. Transparency is key, with insight into why each decision was made adding critical context. This needs to be combined with assessment of the models against fairness criteria with consideration given to any potential harm. Any degree of harm needs a corresponding reproduceable, and reliable mitigation strategies put into continuous operation.
Creating the perfect AI environment
AI has great potential, however there are considerations to avoid common pitfalls. Every AI system comes with its own risks that impact regulatory compliance, such as bias, discrimination and misinformation. These are compounded by data security, scalability and cost management risks. The last three can be resolved with the right AI native architecture.
AI requires a secure, flexible architecture to support its increasingly complex and diverse ecosystem. Many organisations have been on a public cloud journey, however a recent survey discovered that 83% of enterprises plan to move workloads back to private cloud. Public cloud solutions just aren’t right for many applications and workloads, especially AI at scale. They can expose data and models to threats, having significant consequences to intellectual property, copyright and data leakage. Seeking enhanced visibility, increased flexibility and time-saving solutions, many organisations are looking to open source, self-hosted options. However, even these require the ability to govern data usage and monitoring to ensure AI models deliver their intended value.
A hybrid cloud provides a blend of public cloud while providing access with the security and control of an on-premises infrastructure. This means your data can stay within your data centre, preserving data sovereignty whilst still powering your AI applications. A hybrid cloud model with containerisation can streamline application deployment, scaling and management, ensuring consistency and efficiency. This can also help with controlling costs, as running AI solely in the public cloud can be expensive as GenAI is breaking traditional model, with charges for tokens opposed to CPU and GPU usage. A hybrid cloud offers a cost-effective solution by keeping your data on-premises, with consistency of costs. This provides a flexible, scalable infrastructure for AI that truly meets a business need as well as ensuring compliance.
What can HPE do to help?
Having built four of the top ten supercomputers in the world, HPE understands solving challenges at exascale! Something I am passionate about is building AI in a sustainable way, which I have written about extensively , and at HPE we back this up, because six of the ten most energy efficient systems on the Green500 are built by us here HPE. To get started with AI you do not need a supercomputer, but the fundamental requirements of supercomputers and AI workloads are very similar.
At a basic level, a supercomputer is comprised of three parts – interconnected CPUs and GPUs, high speed low latency networking, and a high throughput storage layer. These elements closely resemble the building blocks of AI, with some minor implementation differences.
AI holds immense potential for driving transformation. HPE services provide a flexible, secure and simple approach to harness that power. With a vast and comprehensive AI ecosystem we can make the complexity of AI simple. Where most services focus on day 0 to day 1 challenges, we support day 2 challenges and beyond. With a rich ecosystem of tools to rapidly deploy AI workloads, HPE Private Cloud AI, powered by HPE GreenLake, delivers a self-service cloud experience with a unified control plane, simple infrastructure, and robust tools for rapid AI workload deployment.
Find out more about the potential of AI for FSI here.
Mattab
Matt is Chief Technologist for Artificial Intelligence in the UK&I and has a passion for helping customers understand how AI can be part of a wider digital transformation initiative.