Despite the initial hype around generative AI use cases resembling ChatGPT — for example, a chatbot that answers employees' questions— banks are now evaluating applications based on ROI considerations.
They may not necessarily be the flashiest applications, but the rush to find viable use cases will depend on setting up internal assessment and governance processes to make them happen, experts at Money 20/20 said this week.
"The real question is one of the temporal ordering of use cases," including the readiness of an institution's tech stack and data, and availability of in-house talent, said Prem Natarajan, chief scientist and head of enterprise AI at Capital One. "Everybody thinks they're ready for customer interactions with this. I'm not sure how ready everyone is."
For some companies, the business case for gen AI use cases "under the hood" seem to make sense. Mastercard said it's focusing on safeguarding the transaction environment through gen AI, including combating fraud, said Greg Ulrich, the company's executive vice president and chief AI and data officer.
"We're trying to make the transaction environment safer. How do you improve fraud models?" he said. "How do we make the ecosystem smarter, more intelligent? It's about a recommendation engine helping our partners."
The payments network is also using the technology to improve customer experiences through personalization and working on ways to deploy gen AI to enhance internal operational efficiencies, according to Ulrich. Internal use cases include coding, driving efficiency for engineers and customer support — examples that tap gen AI's capability to make sense of unstructured data.
Similarly, payments firm TSYS plans to use gen AI's capability to fight fraud and cyberattacks, tapping its ability to detect anomalous transactions and perform real-time scoring, said Dondi Black, executive vice president and chief product officer at TSYS.
Firms should be diligent in evaluating the efficiency of testing capabilities, as well as the volume and quality of data, said Natarajan. Firms also need to be able to properly observe and monitor gen AI models.
Build versus buy
Institutions presenting agreed that building everything in-house may not be the most viable option.
"If those tech requirements, which include data, are broadly available in the world, and there's nothing about your data that makes it unique, then there's no reason to build," said Natarajan.
Companies should also evaluate, when making build-versus-buy decisions, whether the solution would be something with which the company intends to differentiate.
"I don't think you build differentiation by becoming a system integrator that integrates three or four different solutions from elsewhere," he said.
Companies may also want to consider how best to deliver privacy assurances to customers.
"There may be other assurances that you want to deliver to your users about their data or the quality of the solutions, or be able to answer questions about these solutions, and to be able to do a deep inspection of those things," said Natarajan.
For Mastercard, it comes down to the sensitivity of the data.
"If there isn't really sensitive information that you're using in there, then we generally try to figure out if there's an existing solution that we can use," said Ulrich.
Governance models
Companies looking to roll out gen AI uses should have a clear governance model to ensure testing parameters are consistent, ethical principles are applied and firms are casting a wide enough net in their consultative efforts.
"It's very important to have a governance framework set up for this very early, and have established protocols about how you're going to be testing for this, how you're going to think about the challenges of deploying gen AI, because otherwise you can run into a lot of problems going through it, and you can trip up on internal processes," said Ulrich.
TSYS set up a center of excellence to establish standards, including data protocols.
"Making sure the data is complete … is not only going to power better output in terms of model performance, it's also going to speak directly to how you inherently keep trust in the model and keep bias out of the model," said Black, who noted that companies need to use AI to continuously retrain models to ensure their effectiveness.
TSYS, in its governance approach, also prioritizes the explainability of AI and how decisions are made, with legal and privacy teams having a seat at the table, she said.
Meanwhile, Mastercard has established an AI and data council chaired by Ulrich and the company's chief privacy officer to ensure all relevant stakeholders — including technologists as well as legal, procurement and business unit heads — are consulted on AI strategies, he said. The group has a particular focus on governance, privacy and bias detection. In turn, employees are kept in the loop about AI's risks and opportunities.
Natarajan of Capital One suggested privacy, ethical considerations and risk management need to be addressed in the beginning of any gen AI rollout and baked into the processes.
"These are not add-on fixes at the end of the implementation cycle. You have to start off in the design phase," he said. Key questions to ask involve the representativeness and completeness of the data, as well as the validation and risk management approaches.
It's also important, he said, to forge relationships with AI researchers at universities that are working on addressing the biggest problems.
He drew attention to the bank's multiyear strategic partnerships with universities. Examples include its role in helping establish the Center for Responsible AI and Decision Making in Finance at the University of Southern California, which was supported by a
"The two biggest risks are rushing in … and walking away, so you have to find the balance," he said.
The decreasing cost of developing gen AI applications will be a boon for companies, said Natarajan.
"Your ongoing cost is really the cost of inference [basically, the ongoing cost of running the model], and there's been at least close to two orders of magnitude reduction in that cost in the last 18 months, thanks to work at Nvidia, and thanks to the work at a lot of other places," he said.
