Preparing your enterprise to leverage AI

Sathish Muthukrishnan is the chief information, data and digital officer at Ally Financial. He has held several senior technology leadership roles throughout his 20-plus-year career. Leading technology in complex, highly regulated businesses has given him a wealth of expertise, which he brings to discussions of artificial intelligence (AI) and its application in financial institutions.

“While AI initiatives are technology-led,” he says, “they must be owned by the whole organization. They must be something that everybody feels part of.” Can you talk about how your company has approached its generative AI journey? Because generative AI is a nascent technology, we needed to really understand it and identify relevant use cases for our customers and our business.

With that in mind, we’ve applied three principles in our use of technology. The first is focusing our initial use cases on our internal customers. This has helped us test and learn about the technology and ensure that it’s using our data in the way we intended. The second principle is a commitment to protecting against the disclosure of personally identifiable information (PII) and other sensitive customer information to the external large language model (LLM). Our AI platform performs the gatekeeping function of removing the PII before we invoke the LLM or the generative AI technology. The third principle is what we call “human in the middle.”

By including a human in the process, we ensure that the technology gives us an expected output. These principles are preparing us well to gain a better understanding of the technology and to protect our data so we can explore use cases for our external customers when the time is right.

What infrastructure do financial institutions need to support generative AI implementation?

Any financial institution looking to leverage AI needs to focus on three key components of infrastructure: the data, the cloud and the network. Understanding your data is critical. Obviously, LLMs perform better with access to an organization’s contextual data, but it’s not just the data. It’s also how and where it’s accessed, where it’s computed and how it’s made available.

Having your data centralized enables easy access and better protection. That leads into the second piece, cloud infrastructure. Centralized data on the cloud with easy access to applications running on the cloud is the ideal combination to explore, experiment with and execute generative AI use cases expeditiously.

The final piece is your network. A modernized network and a software-defined approach to network architecture—i.e., one that’s nimble, more efficient and able to be monitored more effectively—ensure that the flow of data can effectively create impactful experiences for your internal and external customers. Once you have these three components of the tech infrastructure defined and advanced, taking advantage of generative AI becomes more meaningful.

As financial institutions start to implement generative AI, what approaches can they take to protect customer data?

It’s easier to protect the data within your four walls with visibility into who is accessing the data, how they’re using it and for what reason. When data moves to the cloud, you’re now reliant on third parties to protect that data and ensure the right access. This may mean losing a bit of control in protecting the data. To address this, financial institutions can 1.) ensure they have a clear understanding of the data they’re sending to LLMs; 2.) clean out the PII before the data goes into LLMs; and 3.) determine whether the LLM can learn from their data or not.

Ally’s AI platform not only protects against the disclosure of our customers’ PII to the LLMs, it also ensures the tokenized data shared with the LLM is forgotten after every session. As financial institutions leverage generative AI, they’ll need to be thoughtful and find creative and new ways to protect customer data.

How is AI changing data collection in the financial services industry?

To ensure optimal customer experience, companies are already collecting information about what customers like, how they use their services and how they interact with their products, as well as information on product performance. This same data is foundational to taking advantage of generative AI.

At our company, we collect data across all our business verticals to continue to better serve our customers. Centralizing our data, solidifying our cloud infrastructure and upgrading our network have positioned us well for a variety of AI implementations, including generative AI.

Our data collection is not going to change because of how we use AI; it’s how we leverage that data that’s evolving. Our uses of the data could change based on the outcomes, such as uncovering efficiencies and providing new experiences for our customers.

What is one piece of advice you would offer other financial institutions and fintechs as they embark on their AI journey?

While AI initiatives are technology-led, to make a meaningful impact for external and internal customers, those initiatives must be owned by the entire organization. If there’s AI experimentation happening solely within the technology team, that’s not going to push the organization forward. Yes, you can prove out the technology, but if it is not led by your customer-facing businesses and functions, you won’t realize the full potential of this technology.

At Ally, we view harnessing this technology as everybody’s responsibility. That’s why we developed an AI playbook that creates a common understanding of what AI is and how it can be used in the context of Ally being a financial services company. It helps our people answer questions like, “How can I bring an idea to life [using AI]?” and “What steps do I have to take to build on my idea, so it becomes a product for the business?”

We’ve also built out training and tools for people to learn to use generative AI. Involving the whole enterprise has been very beneficial, as many of our AI use cases come from our business and functional groups, with technology acting as an enabler.

What are some practical recommendations for financial services companies on leveraging AI?

First, involve your control partners (risk, legal, compliance, audit and technology risk) from day one. At our company, these groups understand what we’re working on and partner with us to determine the feasibility of proposed use cases. The second recommendation is to get executive-level buy-in. You need to demonstrate to your board, CEO and executive committee members the benefits of using this technology and the impact on end customers. The third element is what I call bringing the organization along. Creating something like an AI playbook is a start, but it’s also important to show people in the enterprise what you’re doing with AI and how they can be part of it.

At Ally, we share with employees some of the AI-driven initiatives we’ve launched. We also bring in external speakers to share how other organizations are using AI and why it’s necessary for us to leverage this technology. So, AI initiatives can’t just happen within the realm of tech. They must be something that everybody feels part of.

Sathish Muthukrishnan is Chief Information, Data and Digital Officer at Ally Financial.

A version of this article appeared in the BAI Special Report: Leveraging AI With Human Capital. You’ll find more AI insights there.