Share with your CDO
Informatica, now operating under Salesforce, is betting that master data management is the missing prerequisite for enterprise AI agents. Dan Hein, Field CTO for Data Foundations at Salesforce, frames the core problem simply: a human spotting a duplicate customer record fixes it once, but an agent without access to a unified data layer returns bad answers hundreds of times per minute. Informatica claims to have run across more than 60 internal systems containing customer data before consolidating them, which is a representative number for any large enterprise.
What this means for your business
The organisations most exposed here are the ones that have already committed to agentic AI workflows but treated data quality as a later problem. If your AI investment is sitting on top of fragmented CRM, ERP, and marketing systems that have never been reconciled, the agent doesn’t just underperform, it industrialises the error. The speed multiplier is real, and it turns a data debt that was manageable when humans were in the loop into an active liability the moment agents start acting autonomously.
Hein’s argument about “trusted context” is stronger than it initially reads. The conventional MDM pitch, that a golden customer record is good for reporting, has always been easy to defer. The agentic framing changes the calculus because agents don’t just read data, they act on it, and probabilistic models can’t self-correct when the inputs are structurally wrong. The processes and governance rules already encoded in MDM systems become, in this framing, the guardrails that make an agent’s behavior predictable enough to put in production. That’s a genuine upgrade in the strategic value of work most enterprises have been treating as plumbing.
The piece is written by a publication that sells into the enterprise technology market, and Hein is pitching Informatica’s product suite, so the case for MDM-before-agents is stated with more confidence than the competitive alternatives warrant. But the underlying logic doesn’t depend on Informatica winning. The fragmentation problem Hein describes, departments building agents independently against different data sources with duplicated governance policies, is a pattern that creates real cost overruns regardless of which vendor you’re using. The decision this reframes isn’t which MDM platform to buy. It’s whether your data governance budget is sized for a world where the cost of dirty data compounds at machine speed, not human speed.
Concept deep-dive: Master Data Management
Master data management is the practice of creating a single, authoritative record for a core business entity, typically a customer, product, or supplier, by pulling data from every system that touches it, resolving conflicts, removing duplicates, and enriching the result. Think of it as a continuously updated source of truth that other systems can query instead of each maintaining their own incomplete version. For AI agents that need to act on customer information, MDM is the difference between a reliable briefing and a contradictory pile of notes.
Based on reporting from Informatica: MDM and the cost of AI’s data blind spots, originally published 2026-10-16 00:29:00.
