(US & Canada) Anju Gupta, VP of Artificial Intelligence at Northwestern Mutual, speaks with Marty Poniatowski, Director at AMD, in a video interview about the organization's approach to GenAI projects, managing the data challenges, the role of analytical engineers, and driving value from unstructured data with GenAI.

Northwestern Mutual is an American financial services mutual organization based in Milwaukee, Wisconsin. Speaking of the organizational approach to GenAI projects, Gupta shares that Northwestern had a good foundation regarding the deployment of AI models. She mentions having a centralized AI team for deploying AI.

Further, with the advent of GenAI, Gupta brought the model governance team together to figure out the prospects with GenAI. She recalls hosting an AI symposium to bring everybody along while also forming an AI council, which involves business leaders, risk leaders, and legal and privacy partners.

The AI council was initially formed to gate-check the use cases. After narrowing down the use cases, the organization created an investment budget for GenAI in 2024 and 2025. Next, the entire organization submitted use cases, which were prioritized by the team or department leaders.

Delving further, Gupta states that the use case prioritization matrix is based on the complexity of the project, the ease of running it, and the value proposition. She adds that once a month, the AI council sits for a healthy debate, and after doing the PoC, her team goes back to the council for production approval.

Gupta notes that this approach with GenAI projects has been successful. She also mentions how the AI and business teams come together to understand how to deliver the project that the user needs.

To ensure that the project is utilized by the user, Gupta relies on “analytic translators,” which is a new role, wherein such translators are forward-looking business leaders. These analytic translators can comprehend business, prompt engineering, and drive the adoption of these technologies.

They are then trained in prompt engineering and are the ones who help the teams understand the algorithm, says Gupta. She adds that the role sits on the cusp of what the AI team delivers and what the business wants.

When asked about managing the data challenge, Gupta affirms that Northwestern Mutual understands its data whereabouts and what it delivers. She maintains that the organization has a golden client record, which is a single source of truth for all clients in one database.

Additionally, the company has a script for what good data should look like, along with a fairly robust data quality and management pipeline. Apart from that, Northwestern Mutual has also built a unified data platform.

Every model that is put into production acquires data from the unified data platform, says Gupta. This secure analytical data platform has enabled the organization to run fast with its AI models and deploy them in business.

Moving forward, Gupta highlights the concept of data products within the organization. While showing success with data products is challenging, she states that Northwestern Mutual has worked to quantify the data products.

For instance, there would be a separate data product for the claims section. Therefore, the data product team focuses on building those assets and defining the taxonomy as well. She says that it is a team sport to work closely with the data team.

Furthermore, Gupta mentions introducing the role of analytical engineers within the company, who are essentially data engineers that are targeted toward the AI workload. They have the engineering capability along with solid data engineering skills, and they work hand in hand with the data scientist.

This shift in workload has led to fast problem-solving for the organization, allowing it to progress faster with GenAI and AI. Emphasizing GenAI, Gupta reflects on how unstructured data can be unlocked and leveraged with GenAI to drive value.

To drive value with unstructured data, Gupta affirms launching Servicing GPT for Northwestern Mutual’s call center into production for a couple hundred Customer Service Representatives. This is being done by utilizing the unstructured data, and while the organization has built pipelines regarding that, it is excited to unleash the unstructured assets.

CDO Magazine appreciates Anju Gupta for sharing her insights with our global data community.

PODCAST | Talent Retention Isn’t About the Dollars — Northwestern Mutual AI VP