How Responsible AI is Shaping the Future of Banking and Finance - with Shub Agarwal of U.S. Bank and USC
🎯 Summary
Podcast Episode Summary: How Responsible AI is Shaping the Future of Banking and Finance
This 21-minute episode features Shubh Agarwal (SVP of Product Management for AI/GenAI at U.S. Bank and USC Professor) discussing the critical intersection of Responsible AI adoption and the impending arrival of Agentic AI within the highly regulated financial services sector. The central theme is that successful, scalable AI implementation hinges on rigorous methodology and governance rather than just the underlying models.
1. Focus Area
The discussion centered on Responsible AI adoption in regulated environments (specifically banking), the transition from Generative AI to Agentic AI, and the strategic imperative of establishing robust AI product creation frameworks. A key assertion is that AI is becoming the new User Experience (UX).
2. Key Technical Insights
- Agentic AI’s Core Function: Agentic AI is defined less as a buzzword and more as a capability that enables step-level automation across workflows, bridging gaps previously requiring human intervention between different software programs.
- Responsible AI Foundation: True responsible AI is rooted in decades of data governance, policies, and rigorous data science practices, not merely marketing or output monitoring. Responsible inputs lead to fewer output headaches.
- AI as the New UX: The future of digital interaction will move beyond buttons and screens toward natural, human-like interfaces enabled by multi-modal Generative AI and Agentic AI, making interactions more personal and authentic.
3. Business/Investment Angle
- Prioritized Impact Areas in Finance: Key areas for agentic AI impact include sales conversion funnels (making interactions more relevant), fraud detection (enabling faster, more precise actions), and security/RPA.
- Adoption Strategy: Companies must adopt a “crawl, walk, run” approach, starting with small, specific workflows to prove real business outcomes before scaling, rather than attempting broad, all-encompassing deployments.
- Human Augmentation, Not Replacement: Agentic AI is intended to automate machine-suited tasks, freeing humans to focus on relationship building, communication, and complex judgment, as it will not replace human judgment or core responsibilities.
4. Notable Companies/People
- Shubh Agarwal: Guest, SVP at U.S. Bank, USC Professor, and author of Successful AI Product Creation: A Nine-Step Framework. His expertise bridges enterprise rigor, academia, and practical product development.
- U.S. Bank / USC: Institutions representing the regulated enterprise environment and the academic rigor applied to teaching AI product management.
- Wiley: Publisher of Shubh Agarwal’s framework book.
5. Future Implications
The industry is moving toward a future where all software will either have AI embedded or will become AI software. Financial institutions must rapidly develop systematic methodologies to move AI concepts from “demo to production” while navigating increasing regulatory complexity. The arrival of agentic AI demands a new set of broader, deeply vetted metrics for measurement and governance.
6. Target Audience
This episode is highly valuable for AI/Tech Professionals, Product Leaders, Enterprise Architects, and Financial Services Executives responsible for strategy, governance, and scaling AI initiatives within regulated industries.
Comprehensive Narrative Summary
The podcast episode provided a deep dive into the practical realities of deploying AI in finance, emphasizing that the current threshold moment for AI is defined by the need for responsible, structured adoption. Shubh Agarwal stressed that the financial industry is cautiously optimistic, prioritizing business value and improved user experience while adhering to strict regulatory duties.
A major point of contention addressed was the nature of Responsible AI. Agarwal firmly argued that responsibility is fundamentally about process and governance—the decades-old discipline of data science—rather than being a marketing buzzword tied to model outputs. Failures often stem from management overriding data scientists’ warnings to rush deployment.
The conversation then pivoted to Agentic AI, which Agarwal views as bringing “great automation” alongside “great responsibility.” He predicts Agentic AI will fundamentally reimagine workflows across industries, making customer experiences more personalized and human-like, solidifying the concept that AI is the new UX. Specific high-impact areas in finance include enhancing sales personalization and dramatically speeding up fraud detection response times.
Agarwal cautioned against the inevitable hype cycle surrounding Agentic AI, clarifying that it automates tasks but will not replace human judgment. He advised against overly broad use cases, especially in risk-averse sectors. The primary headwinds for financial leaders adopting Agentic AI are twofold: first, the difficulty in defining clear, agreed-upon metrics for this new complexity before the previous generation (GenAI) is fully governed; and second, the challenge of identifying use cases that are both valuable and palatable to the industry’s risk appetite.
To overcome these hurdles, Agarwal championed his “method over model” philosophy, detailed in his book. He advocates for a systematic, nine-step framework for AI product creation that integrates ethics, experimentation, and rigorous monitoring from the start. This structured approach is essential for moving successful AI pilots into reliable, enterprise-wide production systems.
🏢 Companies Mentioned
đź’¬ Key Insights
"responsible AI starts with the process, not the model."
"all products that we create, all software that we create will have AI embedded in them, infused in them, or they will all become AI software. That's where the evolution is."
"There are a lot of demos out there, but there are very few production applications that are running."
"it's not about the model, it's about the method."
"there is a methodology that is needed to build AI products... I have seen their teams, their people who understand how to do things in pieces, but that knowledge is not democratized in a way that everyone knows how to build AI products."
"We were in a cycle of defining what the right metrics we need to measure for generative AI, and before we could even define them clearly... we have this agentic AI coming in. The metrics are going to be different, they're going to be broader, they have to be thought through deeply and vetted out."