Watch Why Governance and Clean Data Are Critical for Agentic AI in Retail
Video Summary
In this highlight from Breaking Down Retail Silos: Building the AI-Ready Enterprise, Gary Hawkins asks Lori Schafer about governance in the age of agentic AI. Lori emphasizes that retailers must establish strong guardrails, powered by clean master data and flexible governance rules, to keep pace with rapid innovation.
She explains how AI-native master data platforms allow continuous improvement and how retailers are moving from human approvals toward trust in automated decision-making.
Key Takeaways
Governance is essential as AI agents begin executing tasks on behalf of retailers
AI-native Master Data Management (MDM) ensures flexible hierarchies and attributes that evolve over time
Clean data must come first — before experimenting with agentic AI decision-making
Retailers that don’t move quickly risk falling behind competitors that already have governance and clean data foundations in place
Most companies still require human approvals, but leaders are beginning to trust AI for exception-only governance
Video Transcript
Gary Hawkins: When AI agents are executing tasks on behalf of retailers, it seems like companies need to put guardrails in place. This brings up the topic of governance. What are you seeing in this area? Do retailers understand the need for it, and how are they approaching it?
Lori Schafer: With the pace of technology today, governance is more important than ever. But before you can even think about agentic AI, you need to start with clean master data — those golden records — powered by native AI. Along with that, you need governance rules that are flexible and easy to change.
In the past, master data was static. Hierarchies and attributes didn’t evolve. Today, when AI is built directly into the master data platform, you can continuously improve data quality and ensure it’s clean before letting AI agents act on it.
Right now, the leaders in retail are moving toward this. They’ve invested in clean data and are starting to test agentic AI. But many companies are still at the stage where every AI recommendation requires human approval. For example, we work with a very large grocer where people still review and approve everything.
Over time, as the system learns, companies move to a model where humans only review exceptions. That’s the direction things are heading. But I can’t emphasize this enough: governance starts with getting your data right.