Master data management is having a moment—and generative AI is the reason. Tamr CEO Anthony Deighton says many organizations roll out AI agents and models, then discover the underlying enterprise data is incomplete, duplicated, and trapped in silos. When messy data gets handed to automation, the risks grow fast.
Tamr tackles the problem with models at the core, an approach that grew out of MIT research. Deighton explains why that matters: rules produce a yes/no result, while models can reveal where confidence is lowest—so humans spend time on the few decisions that improve everything downstream. Tamr’s “curator hub” is designed around that idea, funneling expert judgment toward the edge cases with the biggest impact.
In the episode, Deighton describes how this applies in complex environments like healthcare staffing, where matching clinicians to open roles requires stitching together information from hospital systems, credentialing databases, and government identifiers. The goal is a unified, 360-degree view that makes high-stakes decisions more reliable.
Deighton also walks through his path from Siebel to Qlik to Celonis, and why he believes the next decade in data is about improving the data itself.