As an industry built on data, financial services has always been an early adopter of AI technologies. In a recent industry survey, 46% of respondents said AI has improved customer experience, 35% said it has created operational efficiencies, and 20% said it has reduced total cost of ownership. Now, generative AI (gen AI) has supercharged its importance and organisations have begun heavily investing in this technology. IDC estimates that global spending on gen AI will grow from $16 billion in 2023 to $143 billion in 2027, at a compound annual growth rate (CAGR) of 73.8%. The financial services industry is expected to be one of the leading adopters, with a projected spending of $24.6 billion in 2027—accounting for 17.2% of the total gen AI market.
While the possibilities of gen AI and large language models (LLMs) are limitless, there are several data challenges and risks financial executives need to be aware of when implementing AI that generates original content. Access to high-quality source data, strong governance controls and robust security are paramount. According to IDC, “Ultimately, gen AI will be widely adopted only if the data, models and applications that use them are trusted by end users and customers.”
Here’s how a robust data strategy can help your financial services organisation overcome the barriers and blocks to gain a competitive edge with gen AI.
A portfolio of potential
First, let’s look at the current use cases for gen AI. Industry investments so far have focused on three areas:
- Customer experience: Gen AI can help financial services companies differentiate themselves by delivering more efficient, effective customer experiences in an industry known for slower, time-consuming processes. For example, institutions can use chatbots and robo-advisors to answer targeted investment questions and deliver behaviour-driven financial advice in seconds. Insurance agents can use LLMs to analyze administrative data and accelerate claims processing.
- Risk and compliance reporting: Regulatory compliance is increasingly becoming a more complex operational task. According to Thomson Reuters, 73% of risk and compliance teams anticipate an increase in the amount of regulatory information released by regulators and exchanges in the coming year. Gen AI can expedite the process of summarizing, synthesizing and implementing these new regulations. Additionally, the use of synthetic data generated by AI can speed up stress testing and the evaluation and prediction of exposure risks, including those related to fluctuating interest rates and potential defaults.
- Market intelligence and portfolio management: Gen AI can help deduce market sentiment and financial trends by analyzing unstructured data such as filings, reports and news articles. This enables banks and asset managers to prepare for unexpected market shifts and quickly reassess and modify their strategies. Enhanced algorithmic simulations, fueled by extensive forecasting data, can provide more accurate and reliable risk-model recommendations. Gen AI can also synthesize financial data and help investors develop more flexible trading strategies.
Battling the bottlenecks
The data that trains gen AI models, and the technology and infrastructure that supports gen AI, are very important for the success of these use cases. As financial executives integrate gen AI models into their analytics and AI roadmaps, they need to be aware of the issues associated with collecting, storing, and processing the data that feeds the models.
- Data quality and access: Gen AI depends on easy, fast access to first-party data within the organization, direct second-party data from partners, and third-party data from external data providers. However, a significant portion of this data is unstructured, posing challenges in terms of search, cataloguing and analysis. Considering that an estimated 90% of all data is unstructured, those who can tap into this new source of insight will surpass their rivals.
- Data security and governance: Careful governance of model training is essential to reduce the risk of exposing sensitive data, such as personal information and the company’s trade secrets and intellectual property (IP). It’s important to note that the IP rights of AI-generated content will be a subject of ongoing legal, regulatory and policy discussions. Hence, financial services organisations should build an adaptive governance framework for gen AI that helps ensure the quality, reliability and security of the data and models used while protecting data privacy.
- Computing resources: The successful implementation of gen AI hinges on selecting the right language model, trained and fine-tuned on domain-specific and use case-specific data. This could mean training models with trillions of parameters, necessitating scalable storage, memory and high computational power. Having a technology platform that allows for high capacity and flexibility is critical. However, the demand for the processing power and storage space needed to handle massive amounts of structured and unstructured data presents a significant challenge, as these resources are often expensive to acquire and manage. Additionally, recruiting for AI-related roles such as AI data scientists, data engineers and AI product owners remains a hurdle.
- Regulatory compliance: Code generated by gen AI must adhere to sector and industry standards and laws, especially when it uses enterprise data. The SEC has recently proposed new AI rules for broker-dealers and investment advisers to address potential biases and conflicts of interest that may arise from using gen AI. And for organisations operating in the EU, the EU AI Act comes into effect in 2024 and will potentially impact certain consumer-facing AI use cases. Along with AI ethics training and awareness throughout the organisation, it’s important to have a secure and governed platform that protects data privacy and complies with the relevant laws and regulations.
A sound data strategy
To capitalise on gen AI while avoiding the pitfalls, financial services companies need a clear and comprehensive data strategy that includes three key elements:
- The elimination of data barriers: The democratisation of data models enables finance leaders to access and analyze data without relying on data scientists, creating a groundbreaking opportunity. Unlocking this potential requires breaking down data silos across the organisation and with partners and external entities. It also requires opening up direct access to raw and curated data in open formats or from third parties. But financial services organisations that utilise on-premises and legacy solutions will find it challenging to collect, unify and share data.
- A modern data infrastructure: Gen AI requires the convergence of massive amounts of data across systems and clouds, and the seamless provisioning of significant computational resources and storage capacity. A modern data platform can provide the power needed to handle these tasks. It can also offer the managed LLM infrastructure needed to dynamically allocate resources, and to run, tune and build LLMs with open source and third-party models. This ensures the models can be built and operated efficiently and can scale as the data volume increases. A modern platform can also provide built-in AI/ML building blocks that put gen AI and ML into the hands of the entire business, not just AI experts.
- Strong governance and security: Financial organisations need a unified, governed and secure environment that supports end-to-end gen AI development. In 2022, the average data breach cost financial services firms almost $6 million. Implementing gen AI without robust security measures could potentially expose companies to increased security risks. Without proper governance controls, firms expose themselves to the risk of privacy breaches, non-compliance with regulations and reputational damage. Strong governance controls are also needed to ensure that the decisions made by gen AI systems are ethical, fair and in line with the firm’s values and regulatory requirements.
Gen AI offers financial services organisations a multitude of competitive advantages. To leverage this pivotal moment of opportunity, organisations must understand the data challenges and adopt a comprehensive data strategy across their organisation.
To learn about how Snowflake can help your financial services organization use gen AI to power better business outcomes, visit Snowflake’s Financial Services Data Cloud. For more advice for planning and executing an enterprise AI strategy, download our white paper Introducing Generative AI into Financial Services Enterprises.
About the author
Rinesh Patel is the Global Head of Financial Service at Snowflake where he is responsible for developing the financial service value proposition and supporting our customers' transition to an enterprise data strategy in cloud. Rinesh has been in the financial services industry for over 20 years in a variety of banking, consulting and business leadership roles with his most recent experience as Global Head of Cloud Platform at London Stock Exchange Company where he established and led a commercially-minded platform team to develop a modern cloud data platform to deliver new data solutions. Rinesh also held a series of commercial leadership roles including leading the Enterprise Data business where he oversaw performance of a portfolio of solutions comprising market data feeds, analytics solutions and managed services. A thought leader in financial services, Rinesh is an active participant in a variety of industry sponsored initiatives and is a contributor to various publications. Rinesh is based in London where he lives with his wife and two daughters.