AI in 2019 for Lending and Financial Services Industry

The drive to increase efficiency will fuel the financial industry’s continued adoption of AI analytics in 2019.

Companies will improve AI and predictive analytics results by reducing harmful built-in data biases and increasing the diversity of users served by AI. Financial services companies will also continue to automate a broader range of mundane tasks.

Addressing Data Bias in 2019

Data used to train AI models often reflects past biases. It is up to the designers, programmers, and data scientists to identify and remediate these biases. Historically system developers have tended to come from a relatively homogenous group. This does not encourage awareness and sensitivity to historical racial, ethnic or gender biases in training data used for AI modeling.

We do not know if 2019 will be a year in which the representation of women, black and Hispanic workers in STEM computing jobs increases. However, we do believe that a backlash of biases in AI algorithms will create pressure in 2019 for more diversity in “AI teams” to cultivate different viewpoints and experiences.

2019 will also be the year for steady improvement in using technology to identify and address data bias. IBM has identified and is working to remove more than 180 different human biases. This includes biases based on gender, race and ideology, that can creep into AI datasets and erode trust between AI applications and the people who use them.

In 2019 the financial services industry will see more engineering time devoted to testing AI products to ensure they are free from harmful biases before they are released. Bias rating systems are being built to quantify the fairness of AI systems by detecting inconsistencies in decision-making and data that is cognitively or parochially biased.

In 2019 AI teams will spend more time preprocessing training data to apply constraints based on fairness and distortion. For example, a Cornell University analysis of fairness and moral issues in machine learning highlights that most AI algorithms are accurate for the majority of cases…at the expense of mistakes for the smaller protected classes in the training data. Research in 2019 will focus on techniques to make AI system less sensitive to the small differences in input samples that might overstate outcomes.

In 2019 AI will further automate routine tasks

Look for AI to make strides in several areas within the finance industry, including:

Mortgage processing: You’ve seen plenty of ads hawking instant approval of mortgages, while the reality is that the process often takes weeks and can end in rejection. In response, lenders are incorporating more AI support for back-office operations to streamline the underwriting process from weeks to hours. This is important to customers who are competing with cash offers and need a quick commitment.

Replacing OCR: The financial industry processes millions of documents each year, many through optical character recognition. OCR is dumb, in that it doesn’t “know” what to do with the information it harvests. AI systems can replace OCR and react to written content, such as credit applications.

Onshoring credit applications: In the past, mortgage and other credit applications were processed offshore in low-wage countries like Pakistan and India. However, tougher regulations, a better understanding of offshoring costs and rising complaints are fueling the onshoring of mortgage processing. The additional strain on American back-offices is ripe for increased AI support to keep costs and complaints down.

Functional consolidation: One-stop-shopping is easier on consumers and lucrative to service providers. Using AI, service providers such as realtors should be able to improve the shopping experience through a horizontal model of offerings that give consumers centralized access to items such as financing, legal advice, contractors and experts.

If you are planning to make a big data and IT investment, give 2River a call to ensure you get the LIFT in revenue and performance that you are expecting.

 

If you want to start exploring predictive analytics, sign up for a FREE LIFT account now.

Not ready, learn more in our resource section.

Recent Posts

The Archive