Abhishek Trehan, Vice President of Data Engineering & Data Science, JPMorgan Chase & Co.
The vast majority of leaders (84%) in financial services are looking to increase their investment in generative AI (GenAI), according to a 2024 KPMG survey, and many have already done so to automate daily tasks or to upskill their workforce.
However, based on my experience as the vice president of data engineering and data science at JPMorgan Chase & Co., I’ve found that there are many misunderstandings about this emerging technology. Having evaluated many GenAI technologies—such as large language models (LLMs) and generative adversarial networks (GANs)—I want to explain some common challenges and provide some insights about how to best leverage its potential.
First, let’s look at some of the ways that GenAI is being used within financial institutions:
1. Customer Engagement And Personalization
Most people know that financial institutions are deploying AI-driven chatbots to provide real-time, personalized assistance to customers, but GenAI also has the potential to analyze customer data and provide tailored financial advice and product recommendations.
This personalization can help clients make informed decisions aligned with their financial goals but also comes with technical, ethical and fiduciary concerns that will have to be considered before implementing it.
Customers are interested in using this technology for financial insights. A survey by Marqeta’s Consumer Pulse Report indicates that 36% of consumers (sign-up and download required) are interested in using GenAI for personal finance, with the number exceeding 50% for individuals under 50.
2. Fraud Detection And Risk Management
GenAI models can analyze transaction patterns to detect anomalies indicative of fraudulent activities. Visa, for example, has invested in AI technologies to enhance fraud detection capabilities, aiming to reduce operational expenses and improve customer experience.
Fraud detection requires immense amounts of data, but generative models, such as generative adversarial networks (GANs), can create synthetic financial data to train fraud detection systems.
3. Document Automation and Natural Language Understanding (NLU)
GenAI models can quickly summarize lengthy financial reports and legal disclosures, enabling faster decision making and streamlining investment research.
GenAI can also help speed up internal processes. For instance, Morgan Stanley has developed an AI application, “AI @ Morgan Stanley Debrief,” which can summarize video meetings and generate follow-up emails.
Challenges In GenAI Adoption
Looking ahead, the market GenAI in financial services is projected to experience significant growth, with estimates indicating a surge from $1.09 billion in 2023 to over $12 billion by 2023. While the outlook is promising, financial institutions face several challenges in implementing GenAI:
• Data Privacy And Security: Financial institutions handle large amounts of sensitive client data, making compliance with privacy regulations like GDPR and CCPA critical. GenAI models, especially those that generate synthetic data, must be carefully managed to avoid breaches of privacy. Ensuring the protection of sensitive financial data is paramount, and protecting the privacy and security of their data, and their customers’ data, is a top concern for leaders in the banking sector, according. toan SAS report.
• Building Tech Infrastructure: With many companies in the financial sector having long-entrenched tech systems, integrating GenAI solutions with existing infrastructure requires careful planning and investment. GenAI models are also computationally intensive, especially when training large models like GPT or BERT. In practice, focusing on lightweight models or using cloud-based AI solutions can help reduce the financial burden.
• Interpretability And Trust: One major challenge with GenAI adoption is the black-box nature of many models, especially with LLMs. While they offer powerful predictions, they lack transparency in their decision-making process. Adopting explainable AI (XAI) frameworks, such as SHAP and LIME, can significantly improve stakeholder trust and model acceptance.
• Talent Acquisition: There is a growing demand for professionals skilled in AI and machine learning, posing a challenge for institutions to attract and retain such talent. According to EY research, “44% of leaders [in financial services] cited access to skilled resources as a barrier to AI implementation.”
• Bias And Fairness: GenAI models, especially those relying on historical data, can inherit biases present in the training data. In financial applications, such as loan approvals and credit scoring, these biases can lead to unfair decision making. Addressing model fairness can require continuous validation and tuning of the algorithms, coupled with transparency in decision-making processes.
Best Practices For Implementing GenAI In Finance
Successful implementation of GenAI in finance hinges on a few key best practices that not only ensure operational efficiency but also safeguard against the inherent challenges of GenAI. Here are a few to understand:
1. Adopt a responsible AI framework. Ensuring that GenAI models are developed and deployed ethically requires a strong governance framework. This includes setting up internal review boards, monitoring for fairness and bias and adhering to regulatory guidelines.
2. Properly manage data quality and preprocessing. The quality of the data used to train GenAI models directly impacts their performance. Data preprocessing—such as data cleaning, normalization and feature selection—is critical to ensuring the models can generate meaningful insights.
3. Enable continuous model monitoring and feedback loops. GenAI models are dynamic and need continuous updates to remain effective. Establish real-time monitoring systems for models to track their performance and adapt to changes in the market or customer behavior.
With GenAI, financial institutions can enhance productivity, streamline operations and deliver hyper-personalized client experiences. However, successful implementation requires an approach that addresses ethical considerations, data security and transparency.
By investing in responsible GenAI practices, financial firms can lead the way in shaping the future of data-driven finance, creating systems that are both innovative and equitable.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
