National scoring models: less nuanced data, too much manual review

First Hawaiian Bank (FHB), founded in 1858, is Hawai’i’s oldest and largest financial institution. The bank has 48 branches throughout Hawai’i, three in Guam, and one in Saipan, serving one of the most demographically diverse regions in the United States. Almost 50% of the population consists of Asian and Native Pacific Islanders.

Traditional credit scoring methods failed to provide an in-depth analysis with a high level of granularity, limiting FHB’s ability to safely scale. Relying on traditional scoring models required an extensive layering of additional credit policies to mitigate credit risk, which necessitated a high volume of manual reviews (over 90%).

FHB wanted to enhance its underwriting analysis and ability to identify appropriate credit risk but estimated that building an in-house custom scoring model would take 18-24 months, would have ‘blind spots’ resulting from bank-only data, and would delay the execution of competing priorities. For an effective solution, FHB knew it needed to find a trusted technology partner.

Zest AI: accurate, fair, compliant, and efficient

FHB understood that embracing AI was a path forward to improve the accuracy of credit risk assessment. AI could offer a more precise solution and insights that leveraged richer data to increase automation, but it needed to be fair, compliant, and efficient—requiring little to no in-house IT development and minimal data analytics.

Zest AI checked all the boxes:

• Zest AI could provide a customized machine learning model, using tens of millions of data points derived from almost half a million borrowers and credit bureau data representative of the Hawai’i, Guam, and Saipan geographies.­ž
• Zest AI’s underwriting technology could easily integrate with FHB’s loan origination system, eliminating the need for engineering resources and integration costs.­ ž
• FHB’s model would have customizable and comprehensive SR-11-7 model risk management documentation and is compliant with all FCRA, CFPB, and ECOA regulations.

FHB fully launched Zest AI’s underwriting solution for its credit card portfolio in six months, from initial development and testing to production.

Embracing AI for quicker decisions and better outcomes

Within FHB’s first year using AI-automated underwriting technology, most credit card applicants can now receive an instant decision. FHB increased automated decisioning overall, both approvals and declines, to 55%—a 13X increase from 4%. Instant approvals are at 40%, a 9X increase from 4% of total applications prior to launch. This increase in automation allowed underwriters to focus on more complex loan applications rather than repetitive manual processes. FHB has also seen improved credit risk scoring accuracy manifest in delinquency rates. Accounts approved with passable Zest AI scores are significantly outperforming accounts approved with score exceptions by a factor of 4x.

Where there’s AI, there’s opportunity

The impact of AI on FHB’s lending process is clear and is only getting started. While FHB is looking to expand the use of Zest AI underwriting technology into its auto, consumer loans and lines of credit, and SMB portfolios, it is also looking to incorporate LuLu—a customized GenAI lending intelligence companion, into its arsenal of analytic and strategic insights tools. With LuLu, FHB will be able to instantly access competitive industry data, monitor performance, and key lending insights–without dashboards or traditional reporting constraints. There is also potential to integrate Zest Protect—AI fraud detection technology into the lending process, creating a streamlined origination ecosystem.