---
type: "synthesis"
tags: ["synthesis", "operating-model", "organization-design", "tension"]
sources: ["execution"]
id: "cross-operating-model-debate"
sourceVaultSlug: "hbr-seg-execution"
originDay: 8
articleStem: "hbr-seg-execution"
sourceUrl: "(unified vault: 7 sources)"
sourceTitle: "HBR — Firm Ⅱ-C · Execution quality — correct execution of AI"
---
## Three answers to 'who owns AI?'

- **A060 — distribute it.** Reject the single AI hero ([[contrarian-no-single-ai-hero]]); [[claim-every-leader-a-shaper]] — every senior leader must become an [[concept-ai-shapers|AI shaper]], not defer to technical [[concept-ai-architects|architects]] or one CAIO.
- **A089 — centralize via a hub.** ~60% of leaders run an [[concept-ai-center-of-excellence|AI Center of Excellence]] to standardize, govern, and staff (or a federated business-unit variant).
- **A093 — small central enabler + broad decentralized innovation.** Moody's [[concept-generative-intelligence-group|Generative Intelligence Group (GiG)]] vets/delivers/secures while [[concept-decentralized-innovation-at-scale|'14,000 innovators']] experiment; [[contrarian-decentralized-over-siloed-ai]] rejects the siloed AI division.

## The convergent pattern

These aren't as opposed as they look — all three land on **hub-and-spoke / federated**: a light central function for standards, security, and enablement, plus distributed ownership close to the work. A060 supplies the *human* layer (leaders as shapers), A089 the *structural* layer (CoE), A093 the *cultural* layer (everyone innovates, small team enables). The real disagreement is degree of centralization, and it correlates with regulatory pressure: Moody's kept a tight secure perimeter ([[prereq-secure-infrastructure]]) precisely so it could decentralize safely. The shared enemy is the bottlenecked, siloed 'Chief AI Office' that no one else owns. See [[cross-leadership-differentiator]].