---
id: "contrarian-universal-data-set"
type: "contrarian-insight"
source_timestamps: ["§ Effect #2: Duplication and Contradiction"]
source_url: "https://hbr.org/2025/09/dont-let-ai-reinforce-organizational-silos"
source_title: "Don't Let AI Reinforce Organizational Silos"
tags: ["data-strategy", "systems-thinking"]
related: ["concept-purpose-first-approach", "quote-purpose-not-process", "concept-ai-duplication-contradiction"]
challenges: "The technical assumption that data silos and contradictory models are best solved by creating a single, universal enterprise data set."
sources: ["tail2"]
sourceVaultSlug: "hbr-seg-tail2"
originDay: 2
articleStem: "hbr-tail-130-ai-reinforce-silos"
sourceUrl: "https://hbr.org/2025/09/dont-let-ai-reinforce-organizational-silos"
sourceTitle: "Don’t Let AI Reinforce Organizational Silos"
---
# A universal data set is not the fix for contradictory AI models

**Contrarian insight:** A universal data set is *not* the fix for contradictory AI models.

**Challenges the assumption that:** data silos and contradictory models are best solved by creating a single, universal enterprise data set.

When departments reach conflicting conclusions due to siloed data ([[concept-ai-duplication-contradiction]]), the conventional IT reflex is to build a massive, universal data lake or master data management system to force a single source of truth. The authors argue this is the wrong instinct — the fix is not technical (universal data) but strategic: shifting to a [[concept-purpose-first-approach]] (see [[quote-purpose-not-process]]).

**Enrichment counterpoint (important nuance):** Universal data integration is not useless. Enterprise governance sources still stress data quality, shared assets, and governance foundations as necessary prerequisites for scaling AI. The stronger, more defensible position is that **data integration is necessary but not sufficient** — better data alone does not fix misaligned objectives, but misaligned objectives are also not fixable without a workable data foundation. Present the authors' claim as a corrective to *data-only* thinking, not as a dismissal of data governance.