Q1

A deployed AI model begins showing declining accuracy due to new customer behavior patterns. What is this an example of?

A. Data leakage
B. Model drift
C. Feature engineering error
D. Overfitting

Answer: B
Explanation: Changing real-world patterns cause concept drift, leading to reduced performance.

Q2

Before deploying a high-risk AI solution, the MOST critical governance action is:

A. Increase model complexity
B. Conduct fairness and bias assessment
C. Reduce training dataset
D. Automate decision-making fully

Answer: B
Explanation: High-risk AI must undergo bias and fairness testing before deployment.

Q3

Which document ensures transparency in AI model design and limitations?

A. Risk log
B. Model card
C. Sprint backlog
D. KPI report

Answer: B
Explanation: Model cards document purpose, training data, limitations, and bias risks.

Q4

The PRIMARY purpose of AI governance is to:

A. Improve coding speed
B. Reduce infrastructure cost
C. Ensure responsible, compliant AI use
D. Eliminate human oversight

Answer: C
Explanation: Governance ensures ethical, legal, and strategic alignment.

Q5

A credit scoring AI must include human review capability. This is an example of:

A. Automation
B. Human-in-the-loop
C. Data redundancy
D. Model compression

Answer: B
Explanation: High-risk AI requires human override mechanisms.

Q6

Which metric indicates operational stability?

A. Recall
B. Precision
C. Uptime
D. Bias score

Answer: C
Explanation: Uptime measures operational performance.

Q7

Data drift refers to:

A. Change in deployment region
B. Change in input data distribution
C. Model retraining
D. User interface update

Answer: B
Explanation: Data drift occurs when input data distribution shifts.

Q8

The BEST strategy to reduce AI adoption resistance is:

A. Replace employees
B. Transparent communication & training
C. Hide model logic
D. Increase automation

Answer: B
Explanation: Change management reduces resistance.

Q9

AI incident response planning is part of:

A. Data engineering
B. Governance control
C. Feature scaling
D. Model compression

Answer: B

Q10

Which activity belongs to operationalization phase?

A. Data cleaning
B. Model deployment
C. Ideation workshop
D. Brainstorming

Answer: B

Q11

Bias monitoring supports which ECO domain?

A. Data engineering
B. Responsible AI efforts
C. UI design
D. Infrastructure optimization

Answer: B

Q12

Which is NOT part of AI governance?

A. Access control
B. Audit trail
C. Random hyperparameter tuning
D. Risk classification

Answer: C

Q13

Model versioning helps with:

A. Marketing
B. Traceability
C. Data cleaning
D. Cost reduction

Answer: B

Q14

Concept drift occurs when:

A. Code crashes
B. Business context changes
C. Dataset size increases
D. GPU fails

Answer: B

Q15

Which ensures AI transparency?

A. Black-box system
B. Explainability tools
C. Reduced documentation
D. No reporting

Answer: B

Q16

A project manager monitoring ROI after deployment is focusing on:

A. Ethical metrics
B. Business performance
C. Drift detection
D. Model explainability

Answer: B

Q17

Which stakeholder is MOST critical in AI governance approval?

A. Junior developer
B. Ethics or compliance officer
C. Intern
D. UI designer

Answer: B

Q18

Continuous retraining is required because:

A. AI never works
B. Data environments evolve
C. Coding errors occur
D. Hardware degrades

Answer: B

Q19

Role-based access control prevents:

A. Data cleaning
B. Unauthorized model usage
C. Drift
D. Bias

Answer: B

Q20

Operationalizing AI primarily ensures:

A. Research publication
B. Sustainable business impact
C. Academic testing
D. Prototype experimentation

Answer: B

🔥 Advanced Scenario-Based MCQs

CPMAI – Operationalizing & Responsible AI

Q1. High-Risk Deployment Approval

A healthcare AI model predicts patient readmission risk. The model accuracy is 92%, but fairness testing shows slightly higher false negatives for elderly patients. The business sponsor wants immediate deployment.

What should the AI Project Manager do FIRST?

A. Deploy and monitor later
B. Retrain only on elderly data
C. Escalate to AI governance board before deployment
D. Ignore fairness gap since accuracy is high

Answer: C
Explanation: Healthcare is high-risk AI. Fairness gaps require governance review before production deployment.

Q2. Drift Detection

Six months after deployment, an e-commerce recommendation engine shows declining conversion rates. Accuracy metrics remain stable.

What is the MOST likely issue?

A. Hardware failure
B. Concept drift affecting business KPI
C. Data corruption
D. Model overfitting

Answer: B
Explanation: Model accuracy may remain stable while business context changes, indicating concept drift.

Q3. Regulatory Compliance

A financial AI solution operates across multiple countries. A new data privacy regulation restricts cross-border data sharing.

What is the BEST response?

A. Continue current operations
B. Disable AI entirely
C. Review data governance and re-architect data flow
D. Increase model complexity

Answer: C
Explanation: Regulatory change requires governance-aligned architectural adjustments.

Q4. Human-in-the-Loop Decision

An insurance AI auto-approves claims below ₹50,000. A stakeholder proposes removing manual review to speed up processing.

Which is the MOST appropriate consideration?

A. Remove manual review to reduce cost
B. Conduct risk assessment before removing oversight
C. Approve automatically
D. Increase automation regardless of risk

Answer: B
Explanation: Removal of human oversight in financial decisions requires risk re-evaluation.

Q5. Incident Response

An AI chatbot gives misleading medical advice to users.

What should be activated immediately?

A. Marketing team
B. Incident response plan
C. Data science retraining
D. UI redesign

Answer: B
Explanation: Governance requires predefined AI incident response protocols.

Q6. Adoption Resistance

Employees refuse to use an AI decision-support tool, claiming lack of trust.

What is the MOST effective strategy?

A. Mandate usage
B. Replace employees
C. Conduct transparency workshops and explainability demos
D. Hide model logic

Answer: C
Explanation: Trust increases through transparency and education.

Q7. Model Registry Importance

During an audit, the team cannot identify which model version approved certain loans.

This indicates failure in:

A. Feature selection
B. Model registry and traceability
C. Hyperparameter tuning
D. Training data labeling

Answer: B
Explanation: Operational AI requires strict version control and traceability.

Q8. Ethical Escalation

A recruitment AI shows statistically lower selection rates for candidates from certain regions.

What is the FIRST governance action?

A. Deploy anyway
B. Conduct bias impact assessment
C. Hide metrics
D. Increase automation

Answer: B
Explanation: Bias detection must trigger fairness evaluation.

Q9. Performance Monitoring

Which metric BEST indicates operational reliability?

A. Precision
B. Recall
C. Latency and uptime
D. Confusion matrix

Answer: C
Explanation: Operationalization includes infrastructure performance.

Q10. Strategic Alignment

An AI fraud detection system is accurate but increases manual review workload significantly.

What should the project manager evaluate?

A. Remove AI
B. Business value vs operational cost trade-off
C. Ignore workload
D. Retrain randomly

Answer: B
Explanation: Operationalization must balance business outcomes.

Q11. Explainability Requirement

A regulator demands explanation of loan rejection decisions.

What should have been implemented at deployment?

A. Black-box deep learning only
B. Explainability tools (e.g., SHAP/LIME)
C. Higher GPU capacity
D. Larger dataset

Answer: B
Explanation: High-risk AI must ensure explainability.

Q12. Shadow AI Risk

Business users start using external AI tools without governance approval.

This represents:

A. Innovation
B. Shadow AI risk
C. Improved efficiency
D. Model drift

Answer: B
Explanation: Unauthorized AI use bypasses governance and risk controls.

Q13. Continuous Improvement

Which scenario requires retraining?

A. Stable KPIs
B. Significant input data distribution change
C. No regulatory change
D. Hardware upgrade

Answer: B
Explanation: Data drift triggers retraining cycle.

Q14. AI Risk Classification

Which solution is MOST likely categorized as high-risk?

A. Movie recommendation engine
B. Employee scheduling assistant
C. AI-based cancer diagnosis
D. Marketing copy generator

Answer: C
Explanation: Healthcare diagnosis impacts human safety.

Q15. Governance Ownership

Who should sponsor enterprise AI governance?

A. Junior developer
B. Executive leadership
C. Intern
D. UI designer

Answer: B
Explanation: Governance requires executive accountability.

Q16. ROI Monitoring

After deployment, AI improves accuracy but business revenue remains flat.

What should be analyzed?

A. Accuracy only
B. Business KPI alignment
C. GPU utilization
D. Data cleaning process

Answer: B
Explanation: Operational AI success is measured by business impact.

Q17. Ethical Transparency

Publishing limitations and intended use of AI model demonstrates:

A. Model tuning
B. Responsible AI documentation
C. Marketing
D. Infrastructure scaling

Answer: B
Explanation: Transparency builds trust and compliance.

Q18. Security Oversight

An adversarial attack manipulates input data to mislead the AI model.

Which governance area failed?

A. Change management
B. AI security risk management
C. UX design
D. Dataset labeling

Answer: B
Explanation: Operational AI must include adversarial testing and monitoring.

Q19. Scaling AI

A pilot AI project performs well. Leadership wants enterprise-wide rollout.

What is the FIRST step before scaling?

A. Immediate expansion
B. Governance and infrastructure readiness review
C. Marketing announcement
D. Cost reduction

Answer: B
Explanation: Scaling requires governance maturity and infrastructure validation.

Q20. Lifecycle Responsibility

Which statement BEST reflects CPMAI perspective on AI lifecycle?

A. AI project ends after deployment
B. AI governance ends after approval
C. AI requires continuous monitoring and improvement
D. AI only needs technical oversight

Answer: C
Explanation: CPMAI emphasizes lifecycle responsibility beyond deployment.