Denise M Letcher, Chief Data Officer at PNC Financial Services, speaks with Dessa Glasser, Principal at FRG (Financial Risk Group), in a video interview about the importance of data quality with the increased adoption of AI/ML, the need for CDO team to be involved with the model development lifecycle, having governance in place, and the interesting usage of natural language processing in the bank.
The PNC Financial Services Group, Inc. is an American bank holding company and financial services corporation. FRG (Financial Risk Group) is a leading provider of technology and advisory services in the financial risk marketplace.
Letcher begins the conversation by stating the importance of data quality, especially with the increased adoption of AI and ML. She says that feeding good quality data is critical to getting the right predictions out of AI and ML.
Adding on, Letcher maintains that there are numerous articles about firms losing money due to being unable to scale up analytics and run use cases because of poor data quality. Further, considering data quality as the foundation drives organizational focus on basic data infrastructure to maintain quality as the data comes in.
Moving forward, Letcher shares that she partners in different ways to derive the right results. Firstly, she believes that the CDO team must be involved with the model development lifecycle, and ensure that the framework includes checking the source of data.
Citing an instance, Letcher mentions checking whether there is redundant data across multiple sources throughout the firm as it can create bias.
Therefore, she says algorithms are effective only when paired with quality data. As a consequence, a large part of improving data governance deals with auditing and consolidating repetitive sources.
In continuation, Letcher mentions partnering with modelers and being engaged in overall model development wherein there is governance over the data labs. She states that it is imperative to know if the right controls are in place once a model is developed and put into production. Moreover, it needs to be validated regularly as part of the product development lifecycle, she asserts.
When asked about the usage of Natural Language Processing (NLP) in the bank, Letcher shares an example of using NLP on customer calls or requests. All the calls are recorded and converted to text, and then analytics is used to assess the type of requests, satisfaction, and nature of the call subject. She concludes by stating that the usage of NLP can be very interesting.
CDO Magazine appreciates Denise M Letcher for sharing insights with our global community.