While such front-office use cases can yield high-profile wins, they can also create new risks. Appropriate controls should inform initial planning and help minimize the risk of damage to service quality, customer satisfaction and the bank’s brand and reputation. Banks must also recognize that regulators will pay particular attention to customer-facing use cases and those where AI enables automated decisioning.
Where to act now
Banks should look at use cases through the lenses of value creation and risk. In the near term, banks should focus on driving forward the highest value potential opportunities while factoring in the level of risk exposure. The portfolio of AI investments should accelerate broader bank strategic objectives while capitalizing on near-term quick wins that offer clear value with minimal risk. Internally oriented use cases for generating content and automating workflows (e.g., knowledge management) are typically good starting points.
Starting off small and driving quick wins will allow banks to assess their capabilities, recognize key challenges and considerations, and assess current and prospective partnerships or acquisitions to further scale.
Learning from initial quick wins will provide the momentum to move on to higher-value, higher-risk use cases when the organization is ready. It will also set the stage for using GenAI to transform and reinvent business models.
4. Establish a dedicated center of excellence or a control tower approach
All sizes of financial institutions can benefit by standing up a GenAI center of excellence (CoE) to implement early use cases, share knowledge and best practices and develop skills. However, as their GenAI capabilities mature, organizations can go beyond coordinating talent and projects to adopt a “control tower” approach to develop vision and strategy, provide visibility into GenAI adoption across the organization and strengthen governance models.
Where to act now
Larger banks further along in their AI experimentation should establish a control tower function to not only provide direction and vision, but also document a high-level roadmap to achieving the firm’s GenAI goals. Such a roadmap requires a rethink of the value chain and business model, a full assessment of technology architectures and data sets and evaluation of innovation investments. A control tower approach both provides GenAI leadership and coordinates ongoing execution and deployments. It’s critical that the right controls and metrics be put in place, with adjustments being made over time as business outcomes are tracked and needs change.
For smaller and midsize organizations in earlier stages of GenAI adoption, a CoE will suffice as a first step and coordination point for knowledge. Further, a CoE will allow the organization to incrementally improve capabilities, spread best practices, foster knowledge sharing and promote early use cases.
As banks monitor initial use cases and partnerships, they should continually evaluate use cases for scaling up or winding down, as well as assessing which partnerships to consolidate. Banks will also need to decide how the control tower will interact with the different lines of business, and how ownership of use cases, budget, success and governance should be spread or centralized.
5. Establish governance and controls
GenAI introduces new and heightens existing risks in banking operations. While AI governance processes and controls are somewhat similar to those for legacy technologies, new risks require new models and frameworks, both for internal use cases and use of third-party tools.
Organizations must consider when and how employees can leverage GenAI and evaluate the distinct risks of internal and external use cases. GenAI’s impact on operations is another factor. For example, the application of GenAI to lending decisions could lead to biased outcomes based on protected characteristics (e.g., gender or race). The burden of proof rests with banks, meaning they will need to collect evidence to show regulators why applications are denied and that applicants are considered fairly. Even where there are no legal or regulatory boundaries at present, governance models must be designed to promote responsible and ethical use of GenAI.
Where to act now
As a first step, banks should establish guidelines and controls around employee usage of existing, publicly available GenAI tools and models. Those guidelines can be designed to monitor and prevent employees from loading proprietary company information into these models. Additionally, top-of-the-house governance and control frameworks must be established for GenAI development, usage, monitoring and risk management agnostic of individual use cases.
As banks make further investments in GenAI capabilities and develop new use cases, they’ll need to assess the unique challenges and risks associated with the tooling and adjust governance and controls for each individual use case. New use cases will bring new ongoing requirements for testing and assessments for hallucination, bias and other risks.
Taking advantage of the transformational power of GenAI requires a combination of new thinking about a longstanding challenge for banks — how to innovate while keeping the lights on. But banks clearly understand the urgency; a huge majority are already dedicating resources to GenAI.