Our AI Journey Has Been Like a Blast-Off on a Rocket Ship — Fifth Third Bank Director of IT Data and AI Governance

(US & Canada) Jonathan Paul, VP, Director IT Data & AI Governance at Fifth Third Bank, speaks with Peter Geovanes, Chief Innovation and AI Officer, in a video interview about the AI journey at Fifth Third over the last 18 months, the approach to selecting ROI focused AI use cases, the need for AI governance, the importance of collaboration for AI success.

Paul describes the AI journey at Fifth Third over the past 18 months as an exhilarating experience, akin to launching a rocket. The organization began exploring generative AI, driven by a desire to understand its potential and applications. He likens the process to solving a Rubik's Cube, emphasizing the confusion that often accompanies new challenges, particularly when the resources for guidance, like instructional videos, are lacking.

To navigate this complexity, Fifth Third focused on assembling top-tier talent and recruiting data scientists with experience in model development, AI, and machine learning. This team worked alongside those involved in intelligent automation, fostering a collaborative environment for AI innovation. Given the regulatory nature of the industry, Paul highlights the importance of carefully managing and harnessing emerging technologies to kickstart their journey.

For the first year, the focus was on exploration and understanding the capabilities of generative AI. Paul notes that this phase involved strategic leadership to assess the technology's current state and its potential trajectory over the next two years. By showcasing its functionalities in a controlled setting, the team aligned these advancements with relevant business cases.

Paul emphasizes the importance of familiarizing staff with emerging technologies, noting that until individuals actively engage with these innovations, their potential remains somewhat abstract. To bridge this gap, the organization began presenting digital avatars to leadership and teams in learning and development, demonstrating practical applications such as speech-to-text and text-to-speech models. By initiating simple use cases, they were able to illustrate tangible benefits, initially testing around five to ten concepts in a controlled environment.

Following a year of exploration, the focus shifted towards developing a strategic roadmap for integrating generative AI into key areas of the organization. This process refined the initial ideas down to a couple of serious use cases, each evaluated for return on investment (ROI). The objective became clear — to effectively link ROI to a venture capital funding model, ensuring that investments in artificial intelligence align with strategic business goals.

Rather than indiscriminately allocating funds to AI initiatives, the organization plans to channel resources towards projects that demonstrate measurable value, akin to its approach with intelligent automation and lean process improvements. The benefits gained from these AI applications will be reinvested into the business, enabling the development of additional use cases.

In discussing various business use cases, Paul emphasized that the benefits encompass efficiency gains, competitive advantages, and enhanced customer satisfaction — all of which are crucial for their operations. He highlights the importance of placing customers at the center of their strategies, prioritizing ethical AI and responsible deployment as foundational principles.

To enhance customer interactions, the organization has developed a proprietary chatbot named Genie. This tool not only facilitates smoother customer experiences but is also designed to evolve, potentially integrating with larger language models in the future. Paul notes the importance of gathering customer feedback to ensure that these innovations genuinely provide value.

The Genie chatbot operates around the clock, empowering customers to access self-service options as the system becomes more capable. Paul explains that their approach to AI deployment is practical, addressing challenges that traditional automation and lean processes could not overcome.

Another notable application involves improving document processing, particularly for loan remittances that arrive in various unstructured formats. Previously, staff would manually sift through these documents, akin to opening letters by hand. By introducing Optical Character Recognition (OCR) technology, they have significantly streamlined this process.

The integration of large language models allows for the transformation of unstructured documents into a consistent, structured JSON format, which can then be seamlessly integrated back into their systems. Paul underscores the significant reduction in the time required for prototyping, testing, and deploying such solutions, with a recent project taking only six hours to develop from raw letters to structured data ready for use.

Sharing his thoughts on the critical aspect of AI governance, Paul explains that the deployment of AI technologies is deliberately aligned with the maturity of their governance practices. This alignment ensures that the pace of technological advancements does not outstrip the necessary governance measures. He underscores the significance of safe and responsible deployment, where the technology is closely linked to real business use cases, all within a secure environment.

The conversation around governance involves a gradual approach—crawling before walking and then running. Paul points out the challenges traditional IT governance and risk management face in adapting to the complexities of AI. There isn’t a standard playbook to guide organizations, but frameworks and guidelines are emerging to establish effective guardrails.

Paul also mentions the collaborative effort in education, noting the establishment of AI working groups that involve various teams, including Legal, Compliance, Risk, and Technology Strategy. This initiative aims to create a shared understanding and language around AI, enabling the organization to effectively communicate and advocate for AI initiatives—whether they are customer-facing or designed to enhance internal operations.

He concludes by stressing the need for confidence in security controls and architectural integrity before moving forward with AI projects. This confidence is essential for the organization to progress to more ambitious applications of AI.

CDO Magazine appreciates Jonathan Paul for sharing his insights with our global community.