Where AI Can Reduce Risk and Cost in Audit Processes - with Nathaniel Bell of Wells Fargo

Unknown Source May 16, 2025 30 min
artificial-intelligence generative-ai
22 Companies
48 Key Quotes
2 Topics
1 Insights

🎯 Summary

Podcast Episode Summary: Where AI Can Reduce Risk and Cost in Audit Processes - with Nathaniel Bell of Wells Fargo

This 29-minute episode features Nathaniel Bell, Director of Data Management at Wells Fargo, discussing the practical, proven application of AI, particularly deterministic models, within internal audit and risk management workflows at a major financial institution. The conversation emphasizes moving beyond theoretical discussions to focus on efficiency gains, cost reduction, and enhanced decision-making through automation in highly manual and complex banking processes.


1. Focus Area: The primary focus is the practical application of AI (specifically deterministic and simpler automation) in internal audit and risk compliance workflows within the banking sector. Key themes include accelerating information gathering, testing internal controls, identifying risk patterns, and the necessary foundational capabilities (data quality, system integration) required for transitioning to continuous auditing. The discussion also touches upon the evolving roles of human oversight versus advanced AI (Generative and Agentic).

2. Key Technical Insights:

  • Deterministic Models for Efficiency: Simple, deterministic (rules-based) AI deployments are already delivering significant efficiency and cost savings by automating highly manual tasks like form filling and rule checking, even in complex areas.
  • Shifting from Granular to Holistic Risk View: AI’s strength lies in aggregating granular data analysis across multiple services or silos to present high-level, firm-wide risk patterns to executive leadership, something difficult to achieve through traditional manual reviews.
  • Taxonomy Codification: Applying AI/LLMs to codify existing, rule-based taxonomies allows for faster information retrieval and conversational querying, replacing inefficient “double-clicking” through reports to find underlying data.

3. Business/Investment Angle:

  • Cost Savings via Efficiency: The immediate ROI comes from increasing productivity and reducing the time spent on manual data gathering and control testing, even from seemingly small, siloed use cases when aggregated.
  • Risk Aversion Dictates Deployment: Financial services leaders are risk-averse, preferring to deploy AI in important but non-critical areas first to prove value before scaling, leading to potential siloed implementations.
  • The Need for Foundational Capabilities: Successful transition to continuous auditing and broader AI adoption hinges on strong foundational elements like data quality and system integration to ensure data flows across organizational silos.

4. Notable Companies/People:

  • Nathaniel Bell (Wells Fargo): The expert guest, providing real-world perspective from a major financial institution on current AI deployments in audit.
  • Wells Fargo: Used as the primary case study for practical, proven AI implementation in a highly regulated environment.
  • Mindbridge: Mentioned as the sponsor of this special series on AI and financial workflows.

5. Future Implications:

  • Agentic AI Caution: While agentic AI promises significant workflow revolution (e.g., jumping between platforms), widespread adoption in high-stakes regulated environments like banking is still on the horizon, requiring significant trust and governance.
  • Human Role Evolution: The future role of human auditors/analysts will shift from manual data wrangling to higher-level decision-making, acting as “generals” overseeing automated agents, though complete removal of human oversight is unlikely due to accountability issues.
  • Human-in-the-Loop is Mandatory: Due to regulatory and accountability requirements, a human must remain in the loop to confirm decisions, especially in critical processes, contrasting with the potential for “AI agent talking to AI agent” scenarios in less sensitive areas.

6. Target Audience: This episode is highly valuable for Audit and Compliance Leaders, Risk Management Professionals, Data Governance Officers, and Technology Strategists within heavily regulated industries (especially Financial Services) who need practical insights on deploying current-generation AI tools rather than waiting for future, unproven technologies.

🏢 Companies Mentioned

All I unknown
Now C unknown
Now I unknown
Mark Twain unknown
Social Security unknown
But I unknown
Agentic AI unknown
Ed Sullivan unknown
And I unknown
So I unknown
Thanks Matt unknown
Wells Fargo unknown
Data Management unknown
Nathaniel Bell unknown
Emerge AI Research unknown

💬 Key Insights

"We have a much more sophisticated conversation of, okay, we see your vision, but I need the crumbs to get there. I need the trail of crumbs to get to the vision."
Impact Score: 10
"I can't hold an AI model accountable. Yeah, big problem. You can't punish, you can't fire an AI model, you can't punish them, you can't give them administrative leave."
Impact Score: 10
"these use cases are developed in teams, right? Again, at that kind of service level, they're granular and no one, because they're siloed, no one is taking them collectively to say, yeah, these are a bunch of small ones that we could fix, we could deploy. Collectively, it really fixes some problems."
Impact Score: 10
"prevailing wisdom, wisdom is choose an important, but not do-or-die space, something important to the business to deploy this technology. Don't deploy it as a toy..."
Impact Score: 10
"if we can use those models to pull that complex data analysis together, right? And then say, hey, this is what it's we're seeing. This seems to be a problem, not in this particular service, in this particular service, right? Across your services, this is an issue..."
Impact Score: 10
"AI can really benefit, as you said, simple AI doesn't have to be the brand new agentic models that everybody is talking about."
Impact Score: 10

📊 Topics

#artificialintelligence 67 #generativeai 11

🧠 Key Takeaways

💡 be taxed based on these factors that I, that I deployed AI in very simple deterministic use cases to go find

🤖 Processed with true analysis

Generated: October 05, 2025 at 05:08 PM