14. Digital Six Sigma, Governance and Future Directions

Introduction and Context

Six Sigma began as a disciplined approach to reducing variation, defects and process failure. Its central logic—define the problem, measure performance, analyse causes, improve the process and control the gains—remains valid. However, the environments in which improvement takes place have changed substantially. Processes are now executed through enterprise applications, cloud platforms, connected equipment, robotic systems and digital customer interfaces. Data is generated continuously, often at a scale and speed that exceed the capacity of traditional improvement teams.

Digital Six Sigma refers to the integration of Six Sigma thinking with digital technologies, advanced analytics and digitally enabled governance. It does not mean replacing DMAIC with software. Rather, it extends the improvement discipline by using process mining to reveal actual process behaviour, Internet of Things (IoT) devices to capture operational conditions, automation to execute repeatable controls, digital twins to test alternatives, cloud analytics to connect distributed data and artificial intelligence (AI) or machine learning (ML) to identify patterns and support decisions.

The increased use of technology also creates new responsibilities. Poor-quality data can produce precise but misleading conclusions. An automated decision may reproduce historical discrimination. A connected factory may expose safety-critical systems to cyberattack. A predictive model may be difficult to audit or explain. Personal information may be collected beyond what is necessary. Consequently, contemporary Six Sigma professionals must combine statistical reasoning with data governance, cybersecurity, privacy protection, ethical judgement and sustainability thinking.

This chapter examines the architecture, applications and limitations of Digital Six Sigma and considers how governance and professional practice must evolve.

Learning Objectives

After studying this chapter, the learner should be able to:

1. Explain the meaning and distinguishing features of Digital Six Sigma.

2. Evaluate the contributions of process mining, IoT, automation, digital twins, cloud analytics and AI/ML to improvement programmes.

3. Design a governance approach addressing data quality, privacy, cybersecurity, ethics and auditability.

4. Analyse the risks and limitations of data-driven improvement.

5. Apply a digital improvement framework to an organisational case.

6. Assess emerging developments and the changing capabilities required of Six Sigma professionals.

From Traditional Improvement to Digital Improvement Systems

Traditional Six Sigma generally relies on project data assembled from transactional records, sampling plans, measurements and observations. The improvement team defines a critical-to-quality characteristic, establishes an operational definition, measures baseline performance and uses statistical analysis to identify significant causes. This approach remains essential because digital tools do not determine what matters, whether a measurement is valid or whether an observed association represents a causal relationship.

Digital improvement differs in three important ways. First, it can observe processes at a much higher frequency. Instead of examining a sample of orders, an organisation may analyse every order event, timestamp, routing decision and rework occurrence. Second, it can connect process, equipment, customer and environmental data. Third, it can move from periodic analysis to near-real-time detection and intervention.

The digital approach is best understood as a sociotechnical system. Technology supplies sensing, storage, computation and automation; people provide problem definition, domain knowledge, judgement and accountability. A dashboard does not constitute improvement unless it changes decisions or behaviour. Similarly, an accurate prediction does not create value if the organisation cannot act on it.

A useful distinction is between digitisation, digitalisation and digital transformation. Digitisation converts analogue information into digital form, such as scanning a paper record. Digitalisation uses digital information to improve an existing activity, such as automatically routing an electronic service request. Digital transformation redesigns the operating model, for example by integrating customer, service, inventory and predictive maintenance data to provide outcome-based support. Digital Six Sigma may operate at all three levels, but its strategic value is greatest when it supports redesigned processes rather than merely automating inefficient ones.

The Digital Improvement Architecture

A robust Digital Six Sigma system can be viewed as five connected layers.

The process layer contains the work itself: customer journeys, production activities, approvals, logistics, maintenance and service interactions. Improvement begins here, with value, waste, risk and customer requirements.

The sensing and event layer captures observations. Sensors, application logs, barcode scans, machine states, web clicks, inspection results and employee inputs generate events and measurements. The quality of this layer determines the credibility of later analysis.

The data and integration layer stores, cleans, links and governs information. It may include operational databases, data warehouses, data lakes, application programming interfaces and master-data services. Common definitions and traceability are vital when data originates from different systems.

The analytics layer applies descriptive, diagnostic, predictive and prescriptive techniques. Process mining, control charts, statistical models, simulation, optimisation and machine learning may coexist, but each should be selected for a defined decision need.

The action and governance layer converts insight into intervention. It includes workflow changes, automated controls, human decisions, escalation rules, performance monitoring, access controls, audit records and review mechanisms. A model without a controlled action pathway is an experiment, not a management system.

This architecture also clarifies the DMAIC connection. In Define, digital teams identify the decision, customer requirement and process boundary. In Measure, they establish data lineage, operational definitions and measurement-system adequacy. In Analyse, they combine process discovery with statistical and substantive reasoning. In Improve, they test redesigned workflows, automation or model-assisted decisions. In Control, they monitor both process performance and the continuing validity, fairness and security of the digital solution.

Technologies That Extend Six Sigma Practice

Process Mining

Process mining reconstructs actual process flows from event logs. Each case—such as an order, claim, patient episode or invoice—contains events, timestamps and attributes. Process discovery produces a model of observed paths; conformance checking compares actual behaviour with a reference process; performance analysis identifies waiting, rework, deviations and bottlenecks.

Its major contribution is visibility into the difference between prescribed and performed work. A process map created in a workshop may show one approval sequence, whereas event data may reveal repeated loops, informal workarounds and exceptional routes. Process mining therefore strengthens the Measure and Analyse phases. It can also support continuous monitoring through alerts for prolonged cases or unauthorised transitions.

Process mining is not automatically reliable. Incomplete logs, inconsistent case identifiers, missing timestamps and overly detailed models can obscure rather than clarify the process. A technically elegant model may also ignore human work performed outside the information system. Analysts should validate findings with process participants and distinguish correlation from root cause.

Internet of Things and Edge Analytics

IoT connects physical assets, sensors, networks and software. In manufacturing, vibration, temperature, pressure, current and acoustic sensors can provide evidence of equipment condition. In logistics, location and environmental sensors can monitor shipment integrity. In healthcare and facilities management, connected devices can support utilisation and condition monitoring.

IoT enables higher-frequency measurement and condition-based intervention. A control strategy may shift from reacting to breakdowns toward detecting deterioration before failure. Edge analytics processes selected data close to the equipment, reducing latency, bandwidth requirements and exposure of sensitive information. Cloud systems can then aggregate longer-term data for trend analysis and fleet-level learning.

The Six Sigma questions remain fundamental: Is the sensor measuring the intended characteristic? Has it been calibrated? Is the sampling frequency adequate? Are readings affected by drift, interference or environmental conditions? More data does not compensate for a defective measurement system. Sensor governance should include calibration, version control, time synchronisation, ownership and rules for handling missing readings.

Automation and Robotic Process Automation

Automation executes defined activities with limited human intervention. Robotic process automation (RPA) is particularly useful for repetitive, rule-based interactions across applications, such as validating fields, transferring data or generating standard notifications. Workflow automation can enforce approvals, segregate duties and create consistent records.

Automation can reduce variation and transaction time, but it can also institutionalise a bad process. Before automating, the organisation should remove unnecessary steps, clarify decision rules and confirm exception handling. Automation is most suitable where inputs are stable, rules are explicit and errors can be detected. Activities involving ambiguity, empathy, complex judgement or ethical discretion require carefully designed human involvement.

A control plan for automation should specify ownership, access privileges, exception queues, fail-safe behaviour, change approval and performance review. The relevant question is not simply whether a task can be automated, but whether automation improves the customer outcome without transferring risk elsewhere.

Digital Twins and Simulation

A digital twin is a digital representation of a physical asset, process or system that is connected to relevant data and updated over time. A simulation model may test scenarios without real-time connection; a digital twin generally has a continuing relationship with the real-world object. In operations, twins can examine production scheduling, energy use, maintenance policies, warehouse layouts or patient-flow scenarios.

Digital twins support improvement by allowing teams to test changes before implementation. They are especially valuable where physical experiments are expensive, disruptive or unsafe. However, model assumptions, boundary conditions and data quality determine the validity of results. A twin should be calibrated against observed performance and periodically revalidated when equipment, products or operating conditions change. It should not be treated as a perfect replica.

Cloud Analytics and Data Platforms

Cloud analytics provides scalable computing, shared data access and a common platform for distributed teams. It can combine enterprise resource planning data, customer information, machine data and external sources. Data warehouses typically organise structured information for reliable reporting, while data lakes can retain larger volumes and varied formats. Modern architectures may use both, supported by catalogues, metadata and governed pipelines.

The benefits include elastic capacity, faster deployment and cross-functional visibility. Risks include uncontrolled data copies, unclear ownership, vendor dependence, insecure interfaces and costs that are difficult to forecast. Cloud adoption therefore requires identity management, encryption, retention rules, contractual controls, resilience planning and defined responsibilities between the organisation and the provider.

Artificial Intelligence and Machine Learning

AI is a broad term for systems performing tasks associated with human cognitive capabilities. ML uses algorithms that learn patterns from data to classify, predict or recommend. In Six Sigma programmes, ML may be used for predictive maintenance, defect detection through computer vision, demand forecasting, anomaly detection, routing, customer-risk assessment and text analysis.

The value of ML depends on the decision context. A model should be assessed not only by predictive accuracy but also by false-positive and false-negative consequences, stability across groups or conditions, interpretability, latency, operational cost and actionability. A highly complex model that cannot be monitored or challenged may be less appropriate than a simpler model with transparent logic.

Supervised learning requires labelled outcomes; unsupervised learning searches for structure without a predefined label; reinforcement learning selects actions through feedback. These categories do not remove the need for process knowledge. Historical data may encode past errors, unequal treatment or conditions that no longer exist. Model performance can deteriorate through concept drift when customer behaviour, equipment, regulations or product designs change.

Data Quality as a Process Capability Issue

Data quality is often discussed as an information-technology concern, but it is fundamentally a process capability issue. If a field is frequently missing, duplicated, late or incorrectly classified, the process producing that field is unstable. Common dimensions include accuracy, completeness, consistency, timeliness, validity, uniqueness and accessibility.

A data-quality programme should establish critical data elements, owners, definitions and acceptance rules. Data profiling can reveal missingness, outliers and contradictory values. Reconciliation checks can compare systems. Lineage records how data was created, transformed and used. Where feasible, quality indicators should be monitored like process measures, with thresholds, escalation and corrective action.

Analysts must distinguish missing data from zero, an unrecorded event from a non-occurrence and a delayed record from a late process. Imputation may be appropriate for some analytical purposes, but imputed values must be identified and their effects assessed. Data cleaning should not silently change the meaning of the source data.

Governance, Risk and Responsible Use

Digital governance is the system by which authority, accountability, standards and controls are established for technology and data-enabled decisions. Governance should be proportionate to risk. A model recommending a low-risk marketing message requires different controls from one influencing employment, credit, medical treatment, safety or access to essential services.

Cybersecurity and Resilience

Connected improvement systems enlarge the attack surface. Threats may include unauthorised access, ransomware, malicious alteration of sensor readings, model manipulation, data exfiltration and denial of service. Security should be built into design rather than added after deployment. Important measures include least-privilege access, strong authentication, network segmentation, secure software updates, encryption, vulnerability management, backups, incident response and recovery testing.

Operational technology deserves special attention because compromise may affect physical safety and continuity. Availability, integrity and safety may be more important than confidentiality in some industrial contexts. A risk assessment should identify critical assets, dependencies and safe states if systems fail.

Privacy and Data Protection

Privacy concerns arise when digital systems collect information about customers, employees, patients or citizens. Responsible practice requires a clear purpose, lawful basis where applicable, data minimisation, retention limits, access controls and transparency. Sensitive information should not be collected merely because technology makes collection possible. De-identification reduces risk but does not guarantee anonymity, particularly when datasets can be linked.

Privacy impact assessments are useful before deploying monitoring, facial recognition, location tracking or behavioural analytics. Individuals should receive meaningful information about relevant data use, and organisations should provide mechanisms for correction, challenge or redress where decisions materially affect them. Legal requirements vary by jurisdiction, so governance must include applicable data-protection obligations rather than assuming that a single global rule exists.

Algorithmic Bias and Fairness

Bias may enter through historical decisions, unequal data coverage, proxy variables, measurement practices, sampling, labels or deployment context. Removing a sensitive attribute does not necessarily remove discrimination because correlated variables may remain. Fairness is also context-dependent: equal error rates, equal opportunity and equal outcomes are not identical objectives and may conflict.

A responsible assessment examines performance across relevant groups, investigates disparate error patterns, documents trade-offs and involves affected stakeholders. Mitigation may occur before modelling through sampling or feature design, during modelling through constraints, or after modelling through thresholds and human review. Fairness testing should continue after deployment because populations and conditions change.

Explainability, Auditability and Human Accountability

Explainability concerns whether the behaviour of a system can be understood sufficiently for its purpose. A model used for process diagnosis may require feature importance and visual explanation; a high-impact decision may require a clear reason understandable to the affected person. Explanation is not the same as proving causality.

Auditability requires records of data sources, model versions, training procedures, parameters, validation results, approvals, users, outputs and interventions. Reproducibility may be difficult when data streams change, but versioned pipelines and controlled environments improve it. Human oversight must be substantive. A person who can only approve an automated recommendation without time, authority or information to challenge it is not meaningful oversight.

Sustainability

Digital solutions consume energy and materials through sensors, networks, data centres, devices and equipment replacement. Sustainability analysis should therefore include the full life cycle, not only the apparent efficiency benefit. A predictive-maintenance system may reduce scrap and unplanned downtime but increase sensor manufacturing, data transmission and computing demand.

Teams should evaluate energy intensity, equipment life, repairability, e-waste, supplier practices and rebound effects. Efficient models, edge processing, appropriate data retention and renewable-energy sourcing may reduce impact. Sustainability should be treated as a critical-to-quality requirement alongside cost, speed and reliability.

A Governance-Embedded Digital DMAIC Approach

A practical approach integrates digital governance into DMAIC rather than creating a separate compliance exercise.

Define: State the business and customer problem, decision owner, affected stakeholders, expected benefit, risk classification and boundaries. Identify whether technology is actually needed.

Measure: Map data sources and process events; define critical data elements; validate sensors and measurement systems; document consent, access and retention requirements; establish baseline process and model metrics.

Analyse: Combine statistical analysis, process mining, subject-matter expertise and causal reasoning. Test data representativeness, subgroup performance, security threats and unintended consequences. Avoid treating model output as proof of root cause.

Improve: Compare alternative interventions through pilots, simulation or controlled experiments. Design human escalation, privacy safeguards, cybersecurity controls, exception handling and change-management activities. Estimate total cost and sustainability impact.

Control: Monitor process capability, customer outcomes, data quality, model drift, fairness, security events and user workarounds. Maintain version control, audit trails and periodic reapproval. Define conditions under which the model or automation must be suspended.

This approach changes the meaning of control. Control is not merely keeping a mean and variation within limits; it is maintaining the validity, safety and legitimacy of a digitally mediated process.

Illustrative Case: Predictive Quality in a Food-Processing Plant

Background

The following is an illustrative case. A medium-sized food-processing plant experienced variation in package seal integrity. The immediate effects were rework, line stoppages and occasional customer complaints. The plant had inspection records but limited information about conditions preceding a defective seal.

Situation and Problem

The quality team initially proposed installing temperature, pressure and speed sensors on the sealing equipment and training an ML model to predict defects. However, a preliminary review found inconsistent product codes, missing timestamps and different definitions of “defect” across shifts. Some operators adjusted settings manually, but those adjustments were not recorded.

Analysis

The team first used process mapping and event-log analysis to identify the relevant production runs, changeovers, cleaning cycles and inspection events. A measurement-system study showed that the inspection method itself had inconsistent repeatability. Sensors were calibrated and synchronised with the production system. Historical data was then reviewed by product type, machine, shift and material batch.

Analysis suggested that defect risk increased after certain changeovers and during combinations of speed and temperature rather than because of one isolated variable. The team also found that one shift appeared to have a higher defect rate because its operators recorded minor seal irregularities more consistently. Treating raw defect counts as directly comparable would have produced a misleading conclusion.

Approach and Solution

The plant implemented a staged solution. First, it standardised defect definitions and inspection procedures. Second, it introduced a process dashboard showing seal conditions and inspection results. Third, a relatively interpretable predictive model generated a risk alert rather than automatically changing machine settings. Operators received recommended checks, and a supervisor could pause the line when risk exceeded a defined threshold. All alerts, overrides and outcomes were logged.

A pilot compared the digitally supported line with a similar line using the existing control method. The evaluation considered defect rate, false alarms, downtime, operator workload, energy consumption and material waste. Cybersecurity controls separated the sensor network from general office systems, and access rights were assigned according to role.

Outcome

The pilot demonstrated the value of earlier detection, but it also showed that the first model became less reliable after a packaging-material change. Because model monitoring had been included in the control plan, the drift was detected and the model was retrained only after validation. The plant retained human approval for consequential interventions and continued periodic measurement-system checks.

Learning Points

The case illustrates that Digital Six Sigma is not a technology procurement exercise. Reliable definitions, measurement discipline and process knowledge preceded modelling. The model supported decisions but did not replace accountability. Evaluation included operational, human, security and sustainability outcomes rather than predictive accuracy alone. Finally, control required monitoring the model and the process together.

Common Failure Modes and Professional Implications

A frequent mistake is beginning with a technology solution instead of a verified problem. Organisations may purchase sensors, dashboards or AI platforms while failing to identify the customer requirement or decision that must improve. Another failure is automating an unstable process, thereby increasing the speed and scale of error.

Other common problems include confusing correlation with causation, using data collected for one purpose to make high-impact decisions for another, ignoring undocumented work, overlooking subgroup performance, treating a vendor’s model as a black box outside organisational responsibility and measuring adoption rather than outcomes. Excessive dashboards can create attention fragmentation, while excessive alerts can produce alarm fatigue.

Professional practice is consequently becoming more interdisciplinary. A capable practitioner needs process improvement, statistics, data literacy, systems thinking, experimentation, cybersecurity awareness and ethical judgement. Collaboration with data engineers, domain specialists, security professionals, legal advisers and affected users is often necessary. Certification alone cannot substitute for contextual competence.

Emerging Directions

Several developments are likely to influence future improvement programmes. Process intelligence is moving toward continuous detection and recommendation rather than retrospective mapping. Generative AI may assist with documentation, natural-language querying, root-cause hypotheses and training, but its outputs require verification because fluent language does not guarantee factual accuracy. Edge AI may enable local decisions where latency, privacy or connectivity make central processing unsuitable.

Industrial systems are also becoming more autonomous, increasing the importance of safe operating envelopes and human-machine coordination. Digital product passports and traceability systems may support circular-economy objectives. Federated learning may allow organisations to learn from distributed data without centrally transferring all raw information, although it does not eliminate privacy or security risks. Standards for AI management, cybersecurity and digital assurance are developing, but organisations should not wait for perfect regulation before adopting responsible controls.

The future role of the Six Sigma professional is therefore not simply that of a statistical analyst. It is that of a process architect, evidence evaluator and governance participant who can ask whether a digital intervention is valid, valuable, secure, fair, sustainable and controllable.

Best-Practice Principles

Effective Digital Six Sigma programmes generally follow these principles:

• Start with customer value, risk and a decision problem, not with a fashionable technology.
• Establish operational definitions, data ownership, lineage and measurement validity before modelling.
• Use the simplest analytical method that adequately supports the decision.
• Test process performance and model performance across time, conditions and relevant groups.
• Design human escalation, exception handling and safe failure modes.
• Apply privacy, security, fairness and sustainability requirements from project definition onward.
• Maintain version control, audit trails and clear accountability for automated decisions.
• Pilot changes, evaluate unintended effects and scale only after evidence supports deployment.
• Treat control as continuous monitoring of the process, data, technology and outcomes.

Chapter Summary

Digital Six Sigma combines the discipline of Six Sigma with process mining, IoT, automation, digital twins, cloud analytics and AI/ML. Its purpose is not to digitise every activity, but to improve decisions, process capability, responsiveness and control. Digital technologies provide richer observation and faster intervention, while traditional Six Sigma supplies problem definition, measurement discipline, statistical reasoning and structured experimentation.

The chapter emphasised that data quality is a process capability concern and that digital systems must be governed throughout their life cycle. Cybersecurity, privacy, algorithmic fairness, explainability, auditability and sustainability are integral to responsible improvement. A governance-embedded DMAIC approach enables organisations to evaluate both operational results and the legitimacy of digital interventions. Future practitioners will require interdisciplinary capability and must remain accountable for the consequences of technology-enabled decisions.

Key Terms

Digital Six Sigma: Integration of Six Sigma improvement principles with digital technologies and advanced analytics.

Process mining: Analysis of event logs to discover, compare and monitor actual process flows.

Internet of Things (IoT): Network of connected physical devices that sense, communicate and exchange data.

Edge analytics: Processing data near its source rather than sending all data to a central platform.

Robotic process automation: Software-based automation of repetitive, rule-based digital tasks.

Digital twin: A data-connected digital representation of a physical asset, process or system.

Cloud analytics: Use of remotely hosted computing and data services for analysis and decision support.

Machine learning: Algorithms that learn patterns from data to classify, predict or recommend.

Concept drift: Change in the relationship between inputs and outcomes that reduces model validity over time.

Data lineage: Record of a data element’s origin, transformations, movement and use.

Data quality: Fitness of data for use, including accuracy, completeness, consistency and timeliness.

Data governance: Allocation of authority, standards, ownership and controls for data management.

Cybersecurity: Protection of digital systems and information against unauthorised access, disruption or manipulation.

Privacy by design: Incorporation of privacy safeguards into system design and operation from the beginning.

Algorithmic bias: Systematic and unfair distortion in algorithmic outputs or impacts.

Explainability: Ability to provide understandable reasons for a model’s behaviour or output.

Auditability: Capacity to reconstruct, examine and verify decisions, data and system actions.

Human-in-the-loop: Arrangement in which a person retains a defined role in reviewing or authorising automated outputs.

Sustainability: Management of environmental, social and economic impacts across the solution life cycle.

Review Questions

Short Answer Questions

1. What distinguishes digitalisation from digital transformation in a Six Sigma context?

2. How does process mining support the Measure and Analyse phases of DMAIC?

3. Identify four important dimensions of data quality.

4. What is concept drift, and why does it matter in ML-enabled improvement?

5. Why should cybersecurity be considered in operational technology projects?

Long Answer Questions

1. Explain the five-layer digital improvement architecture and show how it connects with DMAIC.

2. Evaluate the benefits and limitations of IoT and digital twins in process improvement.

3. Discuss how privacy, algorithmic fairness and explainability influence the governance of AI-enabled decisions.

4. Explain why automating an unstable process can increase rather than reduce organisational risk.

5. Describe a governance-embedded Digital DMAIC approach for a high-impact business process.

Analytical / Application Questions

1. A bank uses an ML model to prioritise loan applications, but approval rates differ substantially across demographic groups. How should a Six Sigma team investigate the issue and decide whether deployment should continue?

2. A factory’s predictive-maintenance model produces many false alarms after a product redesign. Diagnose possible causes and propose an appropriate control plan.

3. An organisation wants to connect employee-location data, workflow logs and performance dashboards to improve productivity. Design a responsible improvement approach addressing value, privacy, data quality and employee trust.

22. Multiple Choice Questions with Answers

Question 1

Which activity most directly distinguishes process mining from a conventional workshop process map?

A. It records employee opinions about process difficulty

B. It reconstructs process paths from timestamped event logs

C. It replaces the need for operational definitions

D. It guarantees that the discovered process is causal

Correct Answer: B. It reconstructs process paths from timestamped event logs

Question 2

A manufacturing firm installs sensors but has frequent missing and unsynchronised readings. What should occur before training a predictive model?

A. Increase model complexity

B. Remove all unusual observations

C. Validate the measurement system and data quality

D. Automate machine-setting changes immediately

Correct Answer: C. Validate the measurement system and data quality

Question 3

Which situation is most appropriate for robotic process automation?

A. Resolving an emotionally complex customer complaint

B. Making an ambiguous safety judgement

C. Transferring validated fields between systems using stable rules

D. Replacing all managerial accountability

Correct Answer: C. Transferring validated fields between systems using stable rules

Question 4

What is a central governance concern when a model’s historical training data reflects unequal past treatment?

A. The model may reproduce or amplify algorithmic bias

B. Cloud storage will necessarily become unavailable

C. Process capability cannot be measured

D. Encryption will make the model more accurate

Correct Answer: A. The model may reproduce or amplify algorithmic bias

Question 5

In Digital Six Sigma, a digital twin is most useful when the organisation needs to:

A. Test operational scenarios before making costly physical changes

B. Avoid validating any model assumptions

C. Convert every paper document into an image

D. Eliminate all human involvement in decisions

Correct Answer: A. Test operational scenarios before making costly physical changes

Question 6

Which statement best describes meaningful human oversight?

A. A person automatically approves every system output

B. A person has authority, information and time to challenge relevant outputs

C. A person is named as owner but cannot access the model

D. A person reviews only the system’s marketing material

Correct Answer: B. A person has authority, information and time to challenge relevant outputs

Question 7

A model’s accuracy declines after customer behaviour changes. This is an example of:

A. Data lineage

B. Concept drift

C. Edge encryption

D. Process discovery

Correct Answer: B. Concept drift

Question 8

Why is automating an unstable process potentially harmful?

A. Automation always reduces process speed

B. Automation can reproduce and scale existing errors

C. Automation prevents data collection

D. Automation makes governance unnecessary

Correct Answer: B. Automation can reproduce and scale existing errors

Question 9

Which control most directly improves auditability of an AI-enabled process?

A. Deleting prior model versions

B. Recording data sources, model versions, approvals and interventions

C. Allowing unrestricted access to the model

D. Using only verbal operating instructions

Correct Answer: B. Recording data sources, model versions, approvals and interventions

Question 10

Which principle best reflects responsible Digital Six Sigma practice?

A. Select technology first and define value later

B. Judge a model only by predictive accuracy

C. Evaluate value, risk, fairness, security and sustainability throughout the life cycle

D. Treat control as a one-time software installation

Correct Answer: C. Evaluate value, risk, fairness, security and sustainability throughout the life cycle