Question bank
AI Framework for Transformation (AIFT)
41 AI Framework for Transformation (AIFT) questions with the correct answer and a written explanation, organised across 5 exam domains.
- 41 questions
- 5 domains
- Foundation level
certilumQuestion index
Jump straight to any question. Answers and explanations are inline.
AI Ethics, Trust, and Societal Impact
4 questions in this domain
- Question 1Single Choice
In the context of the AIFT framework, which of the following accurately defines bias in AI systems?
- ASystematic unfair deviations from equitable treatmentCorrect answer
- BRandom errors due to insufficient training data
- CIntentional manipulation by malicious actors
- DTendency to prioritize accuracy over fairness
Why this is the answer
Bias is defined as systematic and unfair deviations from equitable treatment arising from data, algorithms, or human choices, not random errors or intentional acts.
- Question 2Multiple Select
Which two statements are true about post-hoc explanations like LIME and SHAP?
- AThey are generated after the model makes a prediction.Correct answer
- BThey provide a complete and faithful representation of the model's reasoning.
- CThey are model-agnostic and can be applied to any machine learning model.Correct answer
- DThey replace the need for intrinsic interpretability.
Why this is the answer
Post-hoc explanations are applied after prediction and are model-agnostic, but they are approximations and not fully faithful.
- Question 3Interactive
What is the purpose of the code shown?
- AA method for data augmentation
- BA technique for feature scaling
- CA differential privacy mechanism that adds Laplace noiseCorrect answer
- DA regularization technique to prevent overfitting
Why this is the answer
The code adds Laplace noise to data, which is a differential privacy technique to protect individual privacy.
- Question 4Single Choice
According to the AIFT framework, which best describes the digital divide?
- AIt is solely about access to internet connectivity.
- BIt includes access, digital literacy, and agency.Correct answer
- CIt can be solved by providing free hardware.
- DIt only affects developing countries.
Why this is the answer
The digital divide encompasses multiple layers: access, digital literacy, and agency, not just internet access or hardware.
AI Frameworks and Governance
11 questions in this domain
- Question 5Single Choice
Which of the following is one of the five OECD AI Principles?
- AInclusive growth, human-centred valuesCorrect answer
- BUnacceptable risk classification
- CConformity assessment
- DCE marking
Why this is the answer
The OECD AI Principles include inclusive growth, human-centred values, transparency, robustness, and accountability.
- Question 6Multiple Select
Which two of the following are risk categories defined in the EU AI Act?
- AUnacceptable riskCorrect answer
- BHigh riskCorrect answer
- CModerate risk
- DNegligible risk
Why this is the answer
The EU AI Act defines four risk categories: unacceptable, high, limited, and minimal.
- Question 7Interactive
Examine the code block showing requirements for an AI management system (AIMS) per ISO/IEC 42001. Which requirement is missing?
- AAI system impact assessmentCorrect answer
- BData encryption
- CModel deployment pipeline
- DUser acceptance testing
Why this is the answer
ISO/IEC 42001 includes AI system impact assessment as a key component of the AIMS.
- Question 8Single Choice
What does Transparency in AI ethics require?
- AOpenness of decision-making processesCorrect answer
- BEqual treatment of all demographic groups
- CAssignment of human responsibility for outcomes
- DProtection of personal data from unauthorised access
Why this is the answer
Transparency mandates that AI decision-making be open to inspection and comprehensible to stakeholders.
- Question 9Multiple Select
Which two practices are used to achieve Fairness in AI?
- ABias identification and mitigationCorrect answer
- BAdversarial debiasingCorrect answer
- CPublishing all training data
- DAssigning a single accountable person
Why this is the answer
Fairness is operationalised through bias detection and mitigation techniques such as adversarial debiasing.
- Question 10Interactive
The code block shows a RACI matrix snippet for AI governance. Which role is missing from the accountability model?
- AInternal AuditorCorrect answer
- BAI Model Owner
- CChief Information Officer
- DExternal Regulator
Why this is the answer
The AIFT framework includes an Internal Auditor role as the third line of defence for independent assurance.
- Question 11Single Choice
What is the first step in AI risk management according to AIFT?
- ARisk identificationCorrect answer
- BRisk assessment
- CRisk mitigation
- DContinuous monitoring
Why this is the answer
The AIFT process begins with risk identification to systematically recognise potential AI-related risks.
- Question 12Multiple Select
Which two are mandatory requirements for high-risk AI systems under the EU AI Act?
- ARisk managementCorrect answer
- BHuman oversightCorrect answer
- CThird-party certification for all systems
- DPublic disclosure of all training data
Why this is the answer
High-risk AI systems require a risk management process and human oversight, among other obligations.
- Question 13Interactive
The code block lists data governance tasks. According to AIFT, which important task is missing?
- AData versioningCorrect answer
- BData encryption
- CModel deployment
- DIncident response
Why this is the answer
AIFT specifies that data versioning is essential for reproducibility and auditability of AI models.
- Question 14Single Choice
Which role is primarily responsible for data quality and lineage in AI governance?
- AData StewardCorrect answer
- BData Owner
- CData Custodian
- DAI Model Owner
Why this is the answer
The Data Steward is responsible for data quality, lineage documentation, and metadata management.
- Question 15Multiple Select
Which two roles belong to the first line of defence in AI governance?
- AAI DevelopersCorrect answer
- BData StewardsCorrect answer
- CAI Risk Manager
- DInternal Audit
Why this is the answer
The first line includes operational roles such as AI developers and data stewards who build and manage AI systems.
AI Lifecycle and Operations
8 questions in this domain
- Question 16Single Choice
What is the first phase of the AI project lifecycle according to the APMG AIFT framework?
- AIdeation and Problem FramingCorrect answer
- BData Acquisition and Preparation
- CModel Development and Training
- DDeployment and Integration
Why this is the answer
The AI project lifecycle begins with ideation and problem framing, where the business problem is clearly articulated and success metrics are defined. This phase ensures that AI is the appropriate solution before any technical work begins.
- Question 17Multiple Select
Which of the following are key activities in the data preparation phase? (Select two)
- AHandling missing valuesCorrect answer
- BTraining the model
- CFeature engineeringCorrect answer
- DValidating model performance
Why this is the answer
Data preparation includes tasks such as handling missing values and feature engineering. Model training and validation occur in later phases.
- Question 18Interactive
Examine the code snippet below. Which type of drift is being monitored?
- AData driftCorrect answer
- BConcept drift
- CModel degradation
- DInfrastructure drift
Why this is the answer
The code compares the distribution of input features between training data and production data, which is typical for detecting data drift. Concept drift would require monitoring the relationship between features and target.
- Question 19Single Choice
What is the primary purpose of a feasibility assessment in the AI lifecycle?
- ATo evaluate data availability, quality, and ethical complianceCorrect answer
- BTo train an initial model prototype
- CTo deploy the model into production
- DTo monitor model performance in real-time
Why this is the answer
The feasibility assessment checks whether sufficient, high-quality data exists and whether it can be ethically acquired. It prevents organisations from investing in AI projects that lack the necessary foundations.
- Question 20Multiple Select
Which two statements correctly describe the validation phase?
- AIt uses a hold-out test set to evaluate model generalisation.Correct answer
- BIt occurs after deployment to monitor performance.
- CIt includes fairness and ethical checks.Correct answer
- DIt involves training the model on all available data.
Why this is the answer
Model validation evaluates the trained model on a separate test set and includes responsible AI reviews for fairness. It happens before deployment, not after.
- Question 21Interactive
Analyse the following versioning metadata. Which component is missing that the AIFT framework would require?
- APointer to training data snapshotCorrect answer
- BModel performance metrics
- CDeployment timestamp
- DModel owner name
Why this is the answer
The AIFT framework mandates that every model version must be linked to its training data snapshot for reproducibility and auditability. The metadata shown includes code and hyperparameters but lacks data provenance.
- Question 22Single Choice
In which scenario is Human-in-the-loop (HITL) most appropriate?
- AWhen incorrect decisions have high legal or ethical consequencesCorrect answer
- BWhen the model has very high accuracy
- CWhen the decision volume is extremely high
- DWhen the organisation lacks monitoring tools
Why this is the answer
HITL is mandatory in high-stakes scenarios where errors can cause harm or violate regulations. It ensures accountability and builds trust, even if it reduces automation speed.
- Question 23Multiple Select
Which two activities are part of AI model monitoring?
- ATracking performance metrics like accuracy and latencyCorrect answer
- BConducting initial model training
- CSetting up alerts for data driftCorrect answer
- DLabeling training data for the first time
Why this is the answer
Model monitoring involves continuous tracking of performance metrics and drift detection. Training and labeling are pre-deployment activities.
AI Strategy and Implementation
10 questions in this domain
- Question 24Single Choice
What is the first step in developing an AI strategy aligned with business objectives?
- ACollect data
- BDefine business objectivesCorrect answer
- CSelect a cloud provider
- DBuild a model prototype
Why this is the answer
Strategy must begin with defining business objectives to ensure alignment; all other steps follow.
- Question 25Multiple Select
Which two elements are essential in an AI business case?
- AExpected benefitsCorrect answer
- BData schema design
- CCost estimatesCorrect answer
- DList of API endpoints
Why this is the answer
A business case must articulate benefits and costs to justify investment.
- Question 26Interactive
Review the SQL code. What data readiness issue does it reveal for an AI project?
- AMissing values in key fieldsCorrect answer
- BDuplicate records
- CInconsistent date formats
- DOut-of-range values
Why this is the answer
Counting NULL stock dates indicates missing data, a readiness issue for AI.
- Question 27Single Choice
What is the primary purpose of a feasibility assessment in AI use case prioritization?
- ATo calculate financial ROI
- BTo determine strategic alignment
- CTo gauge data, tech, and talent readinessCorrect answer
- DTo measure user adoption risks
Why this is the answer
Feasibility covers data, technology, and skills readiness for implementation.
- Question 28Multiple Select
Which two criteria are part of the AIFT use case prioritization model?
- AStrategic alignmentCorrect answer
- BNumber of developers
- CEthical and regulatory riskCorrect answer
- DCode complexity
Why this is the answer
Strategic alignment and ethical risk are two key criteria in the AIFT model.
- Question 29Interactive
Analyze the pseudo-code. Which build/buy principle does it illustrate?
- ABuying an off-the-shelf API
- BBuilding a custom model from scratch
- CHybrid: buying a platform and customizingCorrect answer
- DLicensing a pre-trained model
Why this is the answer
Buying a platform and adding custom training is a hybrid approach.
- Question 30Single Choice
When should an organization choose to build an AI solution according to AIFT?
- AWhen it is a core differentiating capabilityCorrect answer
- BWhen the vendor offers lower cost
- CWhen the project has a tight deadline
- DWhen internal AI skills are low
Why this is the answer
Build is justified for core capabilities that provide competitive advantage.
- Question 31Multiple Select
Which two are key activities in change management for AI adoption?
- AStakeholder mapping and engagementCorrect answer
- BDatabase migration
- CTraining and upskilling programsCorrect answer
- DModel hyperparameter tuning
Why this is the answer
Stakeholder engagement and training are core change management activities.
- Question 32Interactive
Review the communication snippet. Which AI change management principle does it demonstrate?
- ATop-down mandate without engagement
- BTransparent communication addressing concernsCorrect answer
- CTechnical training announcement
- DProject status update
Why this is the answer
Addressing fears of job loss and explaining the 'why' builds trust and reduces resistance.
- Question 33Single Choice
Which type of KPI predicts future AI transformation performance?
- ALeading KPIsCorrect answer
- BLagging KPIs
- CFinancial KPIs
- DOperational KPIs
Why this is the answer
Leading KPIs predict future performance, allowing proactive adjustments.
Artificial Intelligence and Transformation
8 questions in this domain
- Question 34Single Choice
According to the APMG AIFT framework, what is the definition of Artificial Intelligence?
- AThe capability of a machine to imitate intelligent human behaviourCorrect answer
- BA machine that can understand and respond to human emotions
- CA system that possesses consciousness and self-awareness
- DA computer that can outperform humans on any intellectual task
Why this is the answer
The AIFT syllabus explicitly defines AI as 'the capability of a machine to imitate intelligent human behaviour'. The other options introduce concepts (emotion recognition, consciousness, general superintelligence) that are not part of this baseline definition.
- Question 35Multiple Select
Which of the following are examples of historical AI approaches mentioned in the AIFT syllabus? (Select two)
- ASymbolic AICorrect answer
- BDeep LearningCorrect answer
- CFuzzy Logic
- DGenetic Algorithms
Why this is the answer
Symbolic AI and Deep Learning are both explicitly listed as key historical approaches in section 1.1. Fuzzy Logic and Genetic Algorithms are not mentioned as separate paradigms in the provided AIFT content.
- Question 36Interactive
The code block above demonstrates a decision-making process. According to the AIFT framework, which type of AI approach does this code represent?
- ASymbolic AICorrect answer
- BMachine Learning
- CDeep Learning
- DGenerative AI
Why this is the answer
The code uses explicit if-elif-else rules, which is characteristic of Symbolic AI (or rule-based systems). No learning from data is involved, so it does not represent Machine Learning, Deep Learning, or Generative AI.
- Question 37Single Choice
In the AIFT framework, a system that recommends optimal routes to a human driver who makes the final decision is an example of:
- AAutomation
- BAugmentationCorrect answer
- CArtificial Intelligence (autonomous)
- DRobotic Process Automation
Why this is the answer
The human retains decision authority, which is the defining characteristic of Augmentation. Automation would execute the route without human input, while autonomous AI would also act independently.
- Question 38Multiple Select
Which of the following characteristics are true for Automation according to the AIFT framework? (Select two)
- AFollows pre-defined instructionsCorrect answer
- BDoes not learn or adaptCorrect answer
- CRequires large datasets to function
- DHandles ambiguous and novel inputs
Why this is the answer
Automation executes rule-based tasks without learning; it does not need large datasets (that is typical of machine learning) and cannot handle ambiguity.
- Question 39Interactive
The code block above represents which concept from the AIFT framework?
- AAutomationCorrect answer
- BAugmentation
- CArtificial Intelligence
- DMachine Learning
Why this is the answer
The code executes a fixed set of rules (if-else) without learning or human involvement, which is characteristic of Automation.
- Question 40Single Choice
According to the AIFT framework, what is the first step in the AI Transformation Journey?
- AConduct a Proof of Concept
- BSelect a technology vendor
- CAssess organisational readinessCorrect answer
- DDefine the business case
Why this is the answer
The AIFT framework emphasises that a foundational assessment of readiness (data, skills, culture) must precede any tactical activities like PoCs, vendor selection, or detailed business cases.
- Question 41Multiple Select
Which of the following stages are part of the AIFT Transformation Journey? (Select two)
- AProof of ConceptCorrect answer
- BScalable Deployment and IntegrationCorrect answer
- CFinal Product Launch
- DVendor Selection
Why this is the answer
The AIFT journey explicitly includes Proof of Concept and Scalable Deployment & Integration. 'Final Product Launch' is not a defined stage, and vendor selection is an activity, not a journey stage.
Frequently asked questions
Everything about the AI Framework for Transformation (AIFT) question bank and how to use it.