The Dystopian Potential of GenAI in Credit Underwriting: All You Need to Know in One Prompt
Using GenAI in credit underwriting will be much harder to implement than McKinsey thinks, and I can already hear the cries of digital dystopia!
AI is great at credit underwriting, and GenAI’s cousin, “machine learning,” has made billions of credit decisions in China for Ant’s Mybank and Wechat’s WeBank since 2015!
AI is at the core of Ant’s “3-1-0” underwriting process: 3 minutes to apply, 1 second for the answer, and 0 human intervention! So, it’s not AI that is the problem.
GenAI can do something that “machine learning,” as used in China, can’t. It can be used in underwriting to evaluate “unstructured data,” meaning everything you’ve written or had written about you.
This is where things get dystopian very quickly. What if a bank includes all your social media posts or those about you? This makes China’s underwriting look tame by comparison.
Imagine this prompt put to a GenAI underwriting assistant to illustrate the problem fully:
“Did the applicant (X) ever have anything written about them to suggest that they are irresponsible, have not repaid a loan, have a history of drug or alcohol abuse, have links to organized crime, or have extreme points of view?”
👉TAKEAWAYS
Challenges to applying GenAI in credit risk:
🔹 impaired fairness in algorithms that confuse or mislead users
🔹 IP infringements, such as copyright violations or plagiarism
🔹 privacy violations resulting from the use of personal or sensitive information to train models
🔹 the generation of malicious content
🔹 security threads and related vulnerabilities
🔹 performance and explainability issues
🔹 the risk of using proprietary data from third parties
🔹 environmental, social, and governance (ESG) effects, such as increased carbon emissions or workforce disruptions
👊STRAIGHT TALK👊
McKinsey is correct; someday, GenAI will be just another tool deployed in credit underwriting. None of us should doubt this.
McKinsey failed to explain how sensitive this will be, and this leads McKinsey to SEVERELY underestimate the challenge!
In China, Mybank used 3,000 data points to ascertain applicants' creditworthiness. The AI accessed structured data (numerical only) collected from users’ history using the Alipay app.
Yes, these 3000 data points could be seen as invasive, but they were at least limited to data on an app you selected and used. This data use is tame compared to what GenAI is capable of.
GenAI will find every public speeding ticket, DUI, or other infraction and anything written about you in the public sector anywhere, whether it’s true or not.
So, the next time you apply for a loan, will you have to give the bank permission to access your social media account data, or will they help themselves?
GenAI is a great technology, but its use in credit underwriting beyond reading application forms will be fraught with controversy.
More than a few will say it is dystopian.
What do you think?
Join our community by subscribing. You’ll be joining an exciting journey down the rabbit hole to our shared digital future—and you’ll be glad you did!
Subscribing is free, but I am increasingly honored by readers volunteering to opt for a paid subscription to recognize my high-quality writing. Thank you!
If you like what I write, and this newsletter has created value for you, and want to support my independent writing with a paid subscription, you have my heartfelt thanks!
Sponsor Cashless and reach a targeted audience of over 50,000 fintech and CBDC aficionados who would love to know more about what you do!
