---
id: "concept-broken-data-foundation"
type: "concept"
source_timestamps: ["¶1", "§ 1. Create and maintain high-quality data."]
tags: ["data-quality", "ai-failure-modes", "siloed-systems"]
related: ["claim-ai-failure-is-data-failure", "concept-single-instance-data", "action-fix-data-infrastructure"]
definition: "The state of fragmented, inconsistent enterprise data systems that causes AI models to generate untrustworthy outputs, leading to the collapse of AI adoption."
sources: ["tail1"]
sourceVaultSlug: "hbr-seg-tail1"
originDay: 1
articleStem: "hbr-tail-107-lenovo-ai-supply-chain"
sourceUrl: "https://hbr.org/2026/05/how-lenovo-built-an-ai-powered-supply-chain"
sourceTitle: "How Lenovo Built an AI-Powered Supply Chain"
---
# Broken Data Foundation

A **broken data foundation** occurs when an organization attempts to layer artificial intelligence on top of fragmented, siloed data systems. In a typical enterprise, forecasting, logistics, and procurement teams operate on separate data models that do not communicate. When AI is applied to this environment, it inherits the underlying inconsistencies in decision-making across these platforms. Consequently, the AI generates recommendations that contradict the native knowledge of experienced human planners. This leads to a rapid erosion of trust in the system, culminating in the collapse of AI adoption — typically within 18 months (see [[claim-ai-adoption-collapses-18-months]]). The failure is almost always misattributed to the AI technology itself, rather than the flawed data infrastructure feeding it — the essence of [[claim-ai-failure-is-data-failure]].

Robert Handfield crystallizes this in the vault's signature line, [[quote-broken-intelligence]]: *"Build intelligence on a broken data foundation and you get broken intelligence, every single time."*

The prescribed remedy is to establish [[concept-single-instance-data]] — a single, authoritative source of truth — by first executing [[action-fix-data-infrastructure]] and then treating data quality as a permanent discipline ([[action-maintain-data-quality]]). Lenovo's five-year [[concept-digital-transformation-1-0]] is the canonical example of repairing the foundation *before* deploying serious AI.

**Definition:** The state of fragmented, inconsistent enterprise data systems that causes AI models to generate untrustworthy outputs, leading to the collapse of AI adoption.


## Related across articles
- [[claim-broad-data-obscures]]
- [[concept-block-group-resolution]]
- [[prereq-data-infrastructure]]
- [[concept-operational-noise]]