AI in financial services helps speed consumer interaction
By using AI in finance, financial organizations are able to speed processes that otherwise would require more manual input from employees or consumers.
Banks, credit card issuers, and other financial organizations are increasingly relying on AI technology to screen potential customers, intelligently invest, and provide better customer services.
AI in financial services -- a heavily regulated industry -- has had a relatively slow start, but has been growing quickly. Regulations have apparently aided in that growth; fields that are governed by strict guidelines can more easily deploy RPA and rule-based algorithms. They have concrete goals to achieve them, with set paths to take to reach them.
Intuit has also created algorithms that help automate the process of filing taxes, said Michael Radwin, vice president of data science at Intuit.
"Most Americans have to file taxes," he said. "If we can make that easier, it's a win-win."
Most Americans have to file taxes. If we can make that easier, it's a win-win.
Michael RadwinVice president of data science, Intuit
Tax guidelines, like most bureaucratic guidelines, are written in what many dismissively refer to as "legalese," a highly structured form of English. This makes it easier to apply natural language processing (NLP) and quickly read consumer data as they file their taxes, Radwin explained.
A supervised machine learning model Intuit has created can then run on top of that data to automatically scan for potential itemized reductions.
"Most Americans don't benefit from itemizing," Radwin said, and the model, by factoring in decades-worth of consumer information, including zip code, age and occupation, knows with a 99% confidence level if taxpayers likely have more deductions to file.
By applying this bit of AI in financial services, Intuit offers customers a more personalized experience, while aiming toward a faster, more accurate tax filing.
Financial services companies such as Kueski use AI to determine borrower eligibility.
Unlike many credit issuers, Kueski doesn't necessarily factor in credit history when determining eligibility.
"A huge percentage of people are unbanked, and many people are underbanked [in Mexico]," said Jaime Romero, CTO of Kueski. Many people have never had access to a credit card, and have almost no credit history, he said.
Pablo Dávalos, director of data science at the 2012 startup, detailed how the AI in financial services system determines eligibility.
Each potential borrower (now, only Mexican citizens may use the service) fills out a questionnaire that includes basic questions, like age, address and job. A machine learning model is run over the information that checks some of it with external data sources. The model spits out a payment probability and a basic credit risk assessment.
That helps determines whether Kueski should accept a new borrower or make a second assessment. Without going into too many details, Dávalos said that assessment digs into the relationships a customer has with other customers, and flags certain factors for manual examination.
Importantly, the models can draw on social demographic information, which is legal to obtain and use in Mexico, to make decisions.
Compared to the AI in financial services systems larger institutions use, Kueski's is quicker and more agile, as it draws on less data and deals with fewer issues, Dávalos said.
In addition, he said, we "incentivize our data scientists to be very free and creative when creating these models."
Looking ahead
According to a 2017 report by multinational professional services company PwC, financial institutions were expected to use more AI in 2018 and beyond.
"In 2018, firms will likely move toward more advanced 'augmented intelligence,' with tools that help humans make decisions and learn from the interactions," the report noted. "Firms can also look to AI as a way to customize product design and develop predictive analytics to improve outcomes such as reduced accident rates."
