---
id: "claim-traditional-training-metrics-fail"
type: "claim"
source_timestamps: ["§ They Measure Real-World Performance", "¶16"]
tags: ["metrics", "training", "kpis"]
related: ["action-track-human-ai-handoffs", "contrarian-training-hours-are-useless", "quote-measure-what-workers-do", "concept-learning-in-the-flow-of-work"]
confidence: "high"
testable: true
sourceUrl: "https://hbr.org/2026/05/the-best-manufacturers-build-ai-with-workers-not-for-them"
sourceTitle: "The Best Manufacturers Build AI with Workers, Not for Them"
sources: ["adoption"]
sourceVaultSlug: "hbr-seg-adoption"
originDay: 9
articleStem: "hbr-cl-78-build-ai-with-workers"
---
# Traditional Training Metrics Capture Exposure, Not Capability

**Claim:** Traditional training metrics — courses completed, total training hours logged — are fundamentally flawed for AI adoption because they capture only a worker's *exposure* to material, offering zero insight into whether they are actually ready to work effectively alongside AI.

**Reasoning.** Measuring what workers *do* matters more than measuring what management *thinks* they do. The real test of capability is whether workers use new tools confidently and consistently in real operating conditions — which requires shifting to performance indicators based on human-AI interaction, per [[action-track-human-ai-handoffs]]. The one-line directive is [[quote-measure-what-workers-do]].

**Confidence: high. Testable: yes.** This is Pillar 3 ("Measure Real-World Performance") of the [[framework-building-ai-with-workers]] and the formal statement of the contrarian insight [[contrarian-training-hours-are-useless]]. It pairs with [[concept-learning-in-the-flow-of-work]] (the training model) — you cannot fairly measure in-flow capability with classroom-completion metrics.

> **Enrichment caveat.** The *shift* away from static completion metrics is well supported; the *specific* replacements the article names (handoff speed, exception resolution, validation/correction frequency) are author recommendations, not universally standardized KPIs.


## Related across articles
- [[action-measure-trust-factors]]
- [[question-measuring-ai-roi]]