Generative AI (GenAI) is on the radar of organizations in just about every industry, and the financial services sector is no exception. In a recent EY survey, financial services leaders were optimistic about GenAI, with 74% saying their organizations were well positioned to take advantage of GenAI.
However, this GenAI gold rush introduces new levels of risk into financial services companies, especially as organizations attempt to scale tightly controlled pilot projects into production environments across the business.
As initial use cases for GenAI emerge — in areas such as credit risk management, fraud prevention and regulatory compliance — leadership teams must keep both measurements and governance in mind to ensure use cases create business value. They also need to put robust controls in place to help ensure that their organization is deploying AI responsibly. To harness the true power of GenAI, organizations will need to assess use case value and risk to inform a longer-term roadmap.
Here are four steps to consider when moving GenAI initiatives from pilot projects to production.
Get the organization to ‘AI-ready’ mindset
GenAI will impact much of the day-to-day work of every employee and function. The people part of the equation is about how AI will augment, not how AI will replace. Think of GenAI as a copilot, not autopilot. Training is critical, and not just on the technical aspects of using large language models. Employees must be good stewards of how they and their firms use GenAI, making sure they are cognizant of issues of risk and bias when sharing output and inputting corporate data. Basic skills will be required and even new roles. The science and art of prompt engineering, for example, is likely to be a required practice for knowledge workers and perhaps even one of several new roles within the company.
Change management within the organization is important as well. To say the least, there is anxiety around the world over how this technology will impact livelihoods. Companies must pay close attention to allaying fears and focus on GenAI as augmenting human potential vs. replacing human labor.
Identify targets and set goals
Because GenAI models are highly adaptive to many tasks across business lines, it can be challenging to identify where the technology is most useful. Starting with a set of core objectives as part of an AI strategy is a good way to align AI business goals with outcomes. Assigning risk values is also a prudent step in identifying targets and setting goals.
“In general, the first set of GenAI projects our financial services clients are tackling are the ones that are lower risk and often more internal facing, where the output is being consumed by employees vs. by an external customer or client,” said Sameer Gupta, EY Americas Financial Services Organization Analytics Leader. “They also tend to be focused on certain themes, such as improved access to knowledge management, which is a top focus now, but also on projects tied to increasing efficiency and the related ROI.”
To that point, initial goal setting should include determining the right metrics to track progress toward desired outcomes. Establishing KPIs is critical to measuring wins and failures along the way.
“This is the year many GenAI projects will move from pilot to production,” Gupta said. “We’ll start to see results and it will be important to pay attention to metrics and make adjustments where needed.”
Create an AI control tower
CIOs are familiar with the concept of a center of excellence (COE) to develop and manage technology capabilities centrally. But as successful as many COEs are, leadership teams may require a new approach for GenAI. Instead, think of a control tower that includes cross-functional participation to develop a strategy and ensure that resources and budgets are aligned with that strategy. This approach makes sense given the broad set of use cases emerging for GenAI, the need to assess and manage risk centrally, and the challenge to maximize business benefits from the limited resource talent in this space.
“Previously, there were three dominant AI use groups in a financial institution: credit, marketing and fraud. Now, every business unit and functional area is seeking ways to use AI and GenAI,” Gupta said.
GenAI applications may address new use cases, such as using conversational chatbots to improve day-to-day productivity. GenAI tools also may complement traditional AI and advanced analytics capabilities. For example, consider credit underwriting on the commercial side. An organization might still use income and cash flow analysis to determine a borrower’s ability to pay or to service a loan. Applying GenAI to that analysis could improve the ability to extract critical information from financial statements or manage the loan covenants in the credit process.
“This means that an organization can work with existing use cases but with a broader application of AI models both from a decisioning and a productivity standpoint,” Gupta said.
An AI control tower approach enables a systemic approach to AI implementation and maturity while effectively managing risk and increasing business value.
Get your data foundation for AI in order
GenAI-driven outcomes are only as good as the data that goes into the technology. CIOs will need to address what are likely to be longstanding challenges with their data infrastructure in preparation for widescale GenAI deployment. Ongoing issues with outdated data or data silos will significantly inhibit the effectiveness of GenAI initiatives.
“It’s one thing to have low-quality data sitting in some silo being used by an analyst on a piecemeal basis,” Gupta said. “But if you’re now making that data accessible to a larger set of use cases, to a larger set of users, the quality issues and associated risks are multiplied by several factors.”
In addition, GenAI’s ability to ingest large volumes of unstructured data will open up new challenges to maintaining data quality and access.
“Organizations have long focused on the quality of structured data,” said Vidhya Sekhar, EY Americas Financial Services Data and Analytics Leader. “Now, with so much unstructured data coming in, you’ll need new governance policies for cataloging, classifying and using that data with adequate content governance.”
The importance of a strong data foundation, combined with the burgeoning set of use cases for GenAI across the business, puts CIOs in a stronger position to lead a rethinking of traditional business processes, Sekhar said.
“CIOs can play a more prominent role as an enabler for business outcomes,” she said. “They’re not just providing a platform — now they can lead by example in using AI to drive efficiency in their own technology operations and will be the lynchpin in an organization’s AI journey to directly impact or influence the outcomes.”
Checkpoints for production-ready GenAI
GenAI is a transformational technology that opens up a world of possibilities for financial services organizations. But CIOs understand the need to balance GenAI’s potential with the realities of governance, security, data privacy and responsible AI.
“GenAI goes beyond traditional AI. The underlying technology is more complex,” Gupta said. “Setting up the right governance, architecture and testing structure is critical to help all stakeholders gain confidence that the solution is production ready.”
Learn more about how financial services organizations can harness the power of GenAI and other technology to transform and grow at scale and speed.