Lifetime Citizen Portal Access — AI Briefings, Alerts & Unlimited Follows
At OCC summit, banks and researchers say AI and alternative data can broaden credit access but call for shared fairness and explainability frameworks
Loading...
Summary
Capital‑market participants and researchers at the Summit urged cautious use of AI/ML and expanded data sources to underwrite credit‑invisible consumers and small businesses, and proposed industry frameworks for fairness, explainability and data governance as preconditions for scaling pilots.
Technology leaders at the Project Reach summit described both the promise and the risks of using advanced machine learning, generative AI and alternative data to expand credit and scale financial advice.
Celia Karam (Capital One) highlighted practical uses such as friction reduction and personalized advice, and described generative AI as a potential tool to scale service delivery. "Generative AI can unlock the next frontier of inclusion possibilities," she said during the technology panel, while stressing the need to pair innovation with accuracy and fairness frameworks.
Researchers and practitioners urged concrete guardrails. Melissa Coidy (FinRegLab) and other panelists noted that supervised machine‑learning methods are already in production at many institutions, but that safe expansion requires explainability, post‑hoc explanation techniques for adverse actions and better, representative training data. Several speakers advocated an evolving industry framework—built on existing fair‑lending and consumer protection rules—to govern model testing, fairness evaluation and the use of non‑traditional data.
Alternative data pilots described on the stage included repurposing early‑warning/fraud platforms and DDA transaction information to assess cash flows for applicants without traditional credit files. Panelists said the early results were promising: deposit‑based underwriting allowed some lenders to approve applicants previously classified as credit‑invisible, and TA platforms tied to open‑banking feeds reduced processing times for small‑business technical assistance.
Next steps the panel urged: develop a shared framework for model governance and fairness metrics; expand access for smaller banks and CDFIs to vetted AI and data tools; and fund the operational capacity of community intermediaries that will implement outreach and customer engagement.
