Enterprise leaders keep chasing bigger AI models, but the real advantage in fraud detection and risk decisioning may be sitting somewhere else entirely: the data architecture underneath. That is the case Rohit Chauhan, former Executive Vice President of Artificial Intelligence & Fraud Solutions at Mastercard, makes in the second installment of a three-part conversation with Robert Lutton, Vice President of Sandhill Consultants.

Chauhan frames the industry’s rush toward agentic AI as an ambition to build a hundred-story tower without pouring a strong foundation first. A weak data architecture means that tower cannot stand no matter how advanced the AI becomes.

The conversation also unpacks Chauhan’s view that trust is not a model output but an architectural outcome. A casino transaction example shows how connected context can turn a seemingly fraudulent charge into a normal one.

Key themes from this conversation include:

• Data over model power: Better data outperforms a more advanced model more consistently than upgrading the AI alone.
• Trust as architecture, not output: Models lose credibility when they fail on incomplete datasets, and that erosion pushes institutions back toward manual decisioning.
• Context changes the verdict: Isolated transactions can mislead fraud systems, while connected context such as location can reverse a flagged decision entirely.
• Foundational investment as a differentiator: Organizations investing in data infrastructure now will separate themselves from those still leaning on technology alone.

Lutton and Chauhan close by previewing the series’ third and final conversation, which turns to identity graphs as a foundation for authentication and Customer 360.

Did you miss the previous episode?

On the move? Take this conversation with you. If you don’t have time to finish the video, you can stream the complete, uninterrupted audio edition of this episode on the go.