Ever since the advent of AI, there has been a widespread fascination with its potential in the financial industry. However, financial institutions need to move beyond the hype and manage their expectations when it comes to AI implementation. According to Garrett Laird, the director of decision science at Amount, AI can indeed deliver measurable business results, but it is not a miraculous solution for all problems.
One of the primary limitations of AI in the financial sector is its lack of explainability. Laird highlights that the new large language models (LLMs) used in AI are essentially black-box models, making it challenging to understand their decision-making processes. This limitation restricts the potential applications and use cases of AI, especially in highly regulated financial product environments.
However, Laird identifies specific areas where AI can be effectively utilized. Outlier detection and unsupervised learning are two such use cases where AI can provide valuable insights. Despite the limitations, Laird believes that AI, particularly through machine learning, can help financial institutions become more compliant by enabling empirical justifications for decision-making processes.
Data preparation plays a crucial role in training AI models. Laird explains that machine learning operations ensure the inclusion of relevant data from reliable sources while preventing the incorporation of discriminatory information or data from protected classes. While data preparation is often considered tedious, it constitutes a significant portion of the overall machine learning process.
In addition to fraud detection, Laird emphasizes the underrated role of AI in credit decisioning. Credit models require robust governance processes, risk management protocols, and accurate data sets. Laird predicts an ongoing shift towards more customer-specific and product-specific credit underwriting models, enabled through the democratization of machine learning. Although this transformation necessitates significant time, data, and human capital, it has the potential to revolutionize the credit industry.
What are the limitations of AI in the financial sector?
AI faces challenges in explainability, particularly with the use of large language models (LLMs), which operate as black-box models. This restricts their applications in highly regulated financial environments.
Where can AI be effectively utilized in the financial sector?
AI can provide valuable insights in outlier detection and unsupervised learning. Machine learning operations can also help financial institutions become more compliant by empirically justifying decision-making processes.
What role does data preparation play in training AI models?
Data preparation is a critical component of training AI models. It involves ensuring the inclusion of relevant data from reliable sources while preventing the incorporation of discriminatory information or data from protected classes.
How can AI revolutionize credit decisioning?
AI has the potential to enable more customer-specific and product-specific credit underwriting models. This democratization of machine learning can lead to improved credit processes, although it requires significant time, data, and human capital investment.