---
type: "synthesis"
theme: "accountability"
sources: ["governance"]
id: "cross-single-owner-principle"
sourceVaultSlug: "hbr-seg-governance"
originDay: 7
articleStem: "hbr-seg-governance"
sourceUrl: "(unified vault: 8 sources)"
sourceTitle: "HBR — Firm Ⅱ-B · Governance, decision rights, leadership, risk"
---
# The One-Owner Imperative

Every decision framework in the corpus insists accountability land on exactly one identifiable party — and every governance article about AI worries about what happens when it can't.

**In human structures the rule is absolute.** RACI's [[claim-single-accountability]] holds that multiple Accountable parties *inevitably* invite power struggles; the fix is [[concept-arci-framework]] (Accountable first) plus [[action-limit-senior-decisions]] and its contrarian edge [[contrarian-four-decisions-a-year]]. OVIS ([[framework-ovis]]) hard-codes the identical rule: "Owner (O): exactly one person, fully accountable." Both frameworks separate the *owner* from the *blockers* — RACI's Consulted/Informed, OVIS's Veto/Influence — precisely so the one-owner line never blurs.

**AI breaks the rule.** When decisions migrate into models and agents, the single owner dissolves. [[question-ai-accountability-d7]] asks who is liable when an AI no one fully understands makes a catastrophic governance error; [[question-enforcing-ai-fiduciary-duty]] asks who bears liability when an autonomous agent breaches trust — the model developer, the agent-layer developer, the hardware maker, or the user. [[concept-ai-fiduciary-duty]] is the proposed patch: re-attach single accountability to a legal *person* by treating the agent as a fiduciary.

The through-line: the corpus is united that diffuse accountability is fatal (see also [[claim-boards-failing-governance]]), and divided only on whether the owner can ever be non-human. See [[cross-fiduciary-thread]] and [[cross-decentralize-risk-ownership]] for the two directions this pulls.