The transformation imperative: generative AI in wealth and asset management

Generative AI (GenAI), the new poster child of AI applications, promises to deliver superior performance while executing information search, retrieval and synthesis tasks on unstructured content, along with content (e.g., text, image, code) generation capabilities. This ability to process large amounts of information, synthesize within context and auto-generate human-friendly responses has inspired the C-suite to imagine how it will disrupt business value chains and position the enterprise for the future while creating value for all stakeholders, and to have a long-standing impact on society.

We have already seen impacts of GenAI across various industries. Service operations, sales and marketing, legal and risk, and technology represent the main areas that will be affected with this emerging technology. While firms can realize early payoffs by optimizing use cases that focus on operational process and efficiency improvements (including software development), significant value creation will come from personalizing user engagement and improving the customer self-service experience.

In a recent survey conducted by the EY Financial Services practice in August 2023, executive or managing directors from wealth and asset management firms with more than $2 billion in revenue were asked to rank the top three areas where GenAI could have the greatest impact on their organization. Clients indicated use cases across the value chain, with alpha generation and financial advice topping the list, followed by client onboarding, marketing and investment operations. Back-office operations ranked third.

This technology is evolving rapidly, and, to take full advantage of its benefits and manage its inherent risks, financial institutions need to move forward in several key foundational areas:

• Harmonizing the enterprise AI strategy with GenAI is a critical step. Organizations need to further cross-team collaboration across lines of business to define and prioritize use cases, along with quantifying key performance indicators and the return on investment (ROI) impact. Additionally, establishing robust operating model playbooks and reusable frameworks, inclusive of appropriate governance and controls for different risk tiers and technology patterns, will help organizations fast-track the deployment of GenAI models.
• Organizations will also need to invest in the infrastructure to scale capabilities, such as cloud-based platforms, computing resources, software frameworks or vendor partnerships. Most financial services firms have made substantial investments in these areas, but they need to be scaled further to deploy GenAI enterprise applications. Establishing scalable technology patterns with the ability to run on both structured and unstructured data within enterprise data infrastructure, along with the adaptable architecture to support multiple large language model options, remains a key step forward. In addition, developing a robust large language model operations process with enhanced monitoring capability will be critical to operationalizing these models.
• Risk and governance should be among the key initiatives that firms target for investing resources as they embark on their GenAI journey. Given the risks posed by GenAI, such as hallucination in some cases, inherent bias and lack of explainability, organizations need to refresh their enterprise legal, risk and compliance (LRC) policy and framework. While no formal regulatory guidance on GenAI exists, firms should consider benchmarking against existing AI guidance, such as the Federal Reserve SR 11-7 (Guidance on Model Risk Management) standards for now, while accounting for heightened risks from large language model usage, along with the potential to create data or outcomes that could expose the enterprise to legal, reputational and financial risks.
• To that end, financial institutions also need to embark on a major effort to update data standards and upgrade existing data assets to ensure sustainable access to data inputs that are necessary to effectively run GenAI models. These models are data-hungry, and organizations that seek to deploy AI responsibly need to invest in accurate, reliable and sustainable data that is free from noise or bias that may impair the final model outcomes. The current enterprise data management strategy needs to evolve to account for emerging risks (e.g., copyright infringement) with input and the generated data.
• As the capability matures, firms will continue to look for purpose-built GenAI models trained on domain or use case-specific data that delivers better model performance. Firms can choose from a wide variety of closed or open-source large language model options, along with large language model-based vendor products. Firms should explore and define fit-for-use patterns while optimizing for operating costs, risks and performance. Firms will benefit from establishing centers of excellence or AI labs to continuously experiment with different technologies to stay ahead of the curve.
• Workforce readiness is another critical step in the adoption of any transformation technology. To unlock its full potential and maximize the ROI of the GenAI investment, firms can expect to see performance improvements in GenAI solutions by enhancing the quality of prompts. They will benefit by upskilling end users on GenAI with focused training on engineering optimal prompts, as well as building technical capacities to deploy GenAI-enabled applications.