USAA is using a combination of artificial intelligence-powered natural language processing and manual tagging to manage consumer complaints as more companies employ AI-based chat bots.  

USAA automatically and manually categorizes text when using machine learning and AI-driven text recognition and analysis tools, Josh Swenson, director of compliance and enterprise complaint management program, said last month at the CBA Live event in Las Vegas. 

“Those who are managing the complaints do manual tagging, but we also have natural language processing going in addition,” he said. “Using a combination of both factors, we found value in … not having the natural language processing drive everything, because you’re going to have challenges in terms of false positives and false negatives.” 

Manual checks necessary 

In fact, financial institutions must balance their use of AI-based text recognition and analysis technology with manual checks, Chris Willis, partner at law firm Troutman Pepper, said at CBA Live. 

AI-based text processing technology, such as ChatGPT and other bots, “isn’t at the level of maturity, consistency and reliability that it could substitute for human review and decisioning of complaints,” Willis said. “Banks and other financial institutions are not ready to sign up for [AI-driven chat] as their communication to a complaining customer.  

“That kind of technology is not perceived to be ready by the regulating community to take over complaint investigation resolutions,” he said, noting that lenders can explore using AI-based chat to support their operations.  

“For example, using it in a backline analytical sense to compare the consistency or correctness of complaint resolutions as a [quality assurance] function — that might be a good use case to try on a test basis,” Willis added. 

Editor’s note: This article first appeared on Auto Finance News, a sister publication to Bank Automation News.