Governance Mandate, Inventory and Decision Rights Review questions Integrated Summary: Governance Mandate, Inventory and Decision Rights The module integrates define the governance mandate and risk appetite, build and classify an ai system inventory, allocate decision rights and accountability across the lifecycle. Judgment begins by defining purpose, context and affected parties, then separating mandatory constraints from trade-offs. An aggregate rate or supplier promise is insufficient: evidence must match version, population and environment; an owner must be able to act; and a review threshold must be capable of changing the decision. Worked cases show how uncertainty leads to bounded scope, redesign or suspension, and how residual risk depends on a control whose operation has been tested. Professional practice preserves the decision, alternatives and assumptions and revalidates after changes in model, data or exposure.
Review questions Test your understanding with these questions. Marking and saved results are available with a free account.
Question 1 Practice analysis: Which governance record closes the accountability gap? Focus: Define the governance mandate and risk appetite. Delegate the use decision to the supplier that built the component. Delay logging until after rollout so users are not slowed down. Retain the same scope because aggregate accuracy is unchanged. Document authority, evidence limits, suspension rights and a recovery path. Question 2 Practice analysis: Which approval condition is properly bounded? Focus: Build and classify an AI system inventory. Use the current-version sample and expire approval when its assumptions change. Generalize the pilot result to every language because the model is shared. Ignore exposure volume while the error percentage remains stable. Count the presence of a policy as proof that its control operates. Question 3 Practice analysis: Which analysis tests the causal pathway? Focus: Allocate decision rights and accountability across the lifecycle. Map output to action and consequence, then test the weakest pathway. Measure interface response time and infer that stakeholder harm declined. Use one anecdote as proof that the system is safe in all contexts. Remove the non-AI alternative so the proposal appears comparatively stronger. Question 4 Practice analysis: Which exception treatment is proportionate? Focus: Map system context and affected parties. Approve the exception permanently because redesign would delay benefit. Report only the portfolio average to simplify executive discussion. Keep the service because prior investment has already been spent. Time-limit the exception, monitor impact and define closure or escalation. Question 5 Practice analysis: Which oversight design is operationally meaningful? Focus: Analyze harm pathways and benefit claims. Treat a human approval click as meaningful oversight in every workload. Use supplier certification to replace testing of the local use case. Test reviewer information, time, authority and behavior under realistic load. Hide uncertainty ranges so the board receives a single confident number. Question 6 Practice analysis: Which action best preserves an eligibility gate? Focus: Estimate uncertainty and consolidate the risk record. Scale now using the overall rate and revisit exceptions quarterly. Constrain scope, name an owner and set a measurable review trigger. Increase the benefit weight until it offsets the unresolved constraint. Add a reviewer without testing workload, information or suspension authority. Question 7 Practice analysis: Which next step best addresses changed evidence? Focus: Design claims, metrics and an evaluation plan. Keep the old approval because the interface did not visibly change. Accept the supplier statement as sufficient evidence for local operation. Revalidate affected claims, populations and controls before staged release. Move directly to retirement without testing a reversible restricted alternative. Question 8 Practice analysis: Which comparison is defensible before release? Focus: Interpret subgroup performance and fairness trade-offs. Separate the severe-error gate before comparing benefits among eligible choices. Average all harms and benefits into one score before applying any gate. Use adoption as proof that outcomes improved for affected people. Treat missing subgroup observations as zero incidents in the denominator. Question 9 Practice analysis: Which governance record closes the accountability gap? Focus: Test robustness, security and generative-AI risks. Delegate the use decision to the supplier that built the component. Delay logging until after rollout so users are not slowed down. Retain the same scope because aggregate accuracy is unchanged. Document authority, evidence limits, suspension rights and a recovery path. Question 10 Practice analysis: Which approval condition is properly bounded? Focus: Design controls and assess residual risk. Use the current-version sample and expire approval when its assumptions change. Generalize the pilot result to every language because the model is shared. Ignore exposure volume while the error percentage remains stable. Count the presence of a policy as proof that its control operates. Question 11 Practice analysis: Which analysis tests the causal pathway? Focus: Govern suppliers and foundation models. Map output to action and consequence, then test the weakest pathway. Measure interface response time and infer that stakeholder harm declined. Use one anecdote as proof that the system is safe in all contexts. Remove the non-AI alternative so the proposal appears comparatively stronger. Question 12 Practice analysis: Which exception treatment is proportionate? Focus: Operate oversight, contestability and incident response. Approve the exception permanently because redesign would delay benefit. Report only the portfolio average to simplify executive discussion. Keep the service because prior investment has already been spent. Time-limit the exception, monitor impact and define closure or escalation. Question 13 Practice analysis: Which oversight design is operationally meaningful? Focus: Build post-deployment monitoring and detect drift. Treat a human approval click as meaningful oversight in every workload. Use supplier certification to replace testing of the local use case. Test reviewer information, time, authority and behavior under realistic load. Hide uncertainty ranges so the board receives a single confident number. Question 14 Practice analysis: Which action best preserves an eligibility gate? Focus: Control material change and revalidation. Scale now using the overall rate and revisit exceptions quarterly. Constrain scope, name an owner and set a measurable review trigger. Increase the benefit weight until it offsets the unresolved constraint. Add a reviewer without testing workload, information or suspension authority. Question 15 Practice analysis: Which next step best addresses changed evidence? Focus: Challenge portfolio reporting and decide retirement. Keep the old approval because the interface did not visibly change. Accept the supplier statement as sufficient evidence for local operation. Revalidate affected claims, populations and controls before staged release. Move directly to retirement without testing a reversible restricted alternative. Create a free account to mark your answers and save your progress.