Governance Mandate, Inventory and Decision Rights · Lesson 3
Allocate decision rights and accountability across the lifecycle
10 min read
Lesson objectives
Analyze separating execution, approval, independent challenge and incident response without creating an accountability gap and justify an evidence-proportionate decision.
Reading tools
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
Decision rights identify who may issue a binding judgment, not who performs the largest number of tasks. Accountability design therefore begins with a catalogue of consequential decisions rather than a generic organization chart. List the judgments that actually change exposure: accepting a use case, opening a data source, selecting a supplier, approving test results, widening the user population, granting an exception, suspending operation, notifying affected parties, and restoring a service after an incident. For each judgment, name the authority, the person who prepares the recommendation, the independent challenger, the operator who carries out the action, and the people who must be informed. One person may hold several roles for a low-impact reversible decision. Combining evidence creation with acceptance of material residual risk requires a stated reason and a compensating check. Responsibility and accountability are different. An operator is responsible for a task; the final decision owner must justify the outcome within delegated authority. Consultation does not automatically grant veto power, and notification after an event is not prior approval. Record the source, type, and limits of authority, including financial, operational, and time boundaries. A manager authorized to approve a confined trial cannot infer authority for population-wide release or acceptance of an irreversible consequence. Connect each decision to the condition that activates it: a threshold breach, changed purpose, control failure, challenge request, or new evidence. Nominate a substitute when the primary authority is unavailable; a directory of names without cover arrangements creates a gap precisely when urgency is greatest. Define the supplier boundary as well. A provider may submit security evidence or deliver a remedy, but it cannot accept on the institution’s behalf the consequences imposed on customers, workers, or other affected people. A defensible accountability record shows who decided, what evidence was available, which options were rejected, when the authorization expires, and who may reopen it. This allows reviewers to examine a reasonable decision that later produced a poor outcome without reducing governance to a search for an individual to blame.
Build the authority protocol in four linked layers. The competence layer attaches each judgment to the institutional function that already holds the mandate, such as service ownership, information security, risk acceptance, data stewardship, or an executive committee. The trigger layer states when authority changes hands by using impact severity, exposure count, data sensitivity, financial limits, or inability to reverse an action. The separation layer addresses a specific conflict: a model developer should not be the sole approver of a release test that the same developer designed, and an operations team should not close a rights-related incident before the competent specialist has examined it. The continuity layer names alternates, response times, communication channels, and a fallback record if normal tools are unavailable. Test this protocol with events, not ideal meetings. Ask what happens at two in the morning, who can disable a feature within ten minutes, how conflicting indicators are resolved, and whether an independent reviewer can inspect the sample rather than trust a team summary. A decision card should contain a decision code, authority, permitted scope, mandatory inputs, confidence boundary, available actions, validity period, and escalation rule. Unlike a conventional RACI grid, it centres on a specific judgment and describes the power attached to each role instead of assigning broad letters to an activity. Examine the refusal pathway too. Can a challenger temporarily stop progression? Where does unresolved disagreement go? Which conditions apply while review is pending? Without explicit answers, “challenge” becomes advice that can be ignored. Finally, test workload. If one committee must approve every minor reversible adjustment, queues and unofficial bypasses will emerge. Delegate recoverable choices inside measurable limits and reserve severe or irreversible decisions for a higher authority. Indicators of effective design include elapsed decision time, percentage of judgments made within delegation, overdue escalations, expired exceptions not closed, and incidents in which personnel could not identify the stop authority.
Lesson summary
The lesson connects separating execution, approval, independent challenge and incident response without creating an accountability gap to a measure, owner and review condition, recognizing that a matrix cannot create authority; it must match actual delegation and organizational structure.