13. Digital Auditing, Industry 4.0, and Artificial IntelligenceIntroduction and ContextQuality auditing has traditionally relied on planned site visits, interviews, document reviews, observation of operations, and sampling of records. Digital transformation is changing each of these activities. Manufacturing and service organizations now generate continuous streams of information through enterprise systems, sensors, cloud applications, mobile devices, robotic equipment, and customer platforms. Consequently, the auditor increasingly examines not only completed records but also the systems that create, control, analyse, and preserve those records. Digital auditing refers to the use of digital technologies, data analytics, remote collaboration, automated evidence collection, and electronically controlled audit workflows to plan, conduct, document, and follow up audits. It does not mean that an audit is simply performed through video conferencing. A digitally mature audit evaluates the reliability of data flows, system controls, algorithms, access rights, audit trails, and automated decisions alongside conventional evidence. Industry 4.0 intensifies this development. Cyber-physical production systems, industrial Internet of Things (IoT) devices, cloud computing, digital twins, advanced robotics, artificial intelligence (AI), and integrated enterprise platforms connect previously separate processes. These technologies can improve visibility and responsiveness, but they also introduce new risks. A dashboard may display inaccurate information; a sensor may be poorly calibrated; an algorithm may reproduce bias; and an apparently immutable blockchain record may preserve incorrect data permanently. For quality professionals, the central issue is therefore not whether technology is being used, but whether digital systems produce trustworthy, relevant, complete, timely, secure, and explainable evidence. The auditor must combine knowledge of quality management systems, information systems, cybersecurity, data governance, process analysis, and professional judgement. Learning ObjectivesAfter studying this chapter, the learner should be able to: 1. Explain the meaning and scope of digital auditing in an Industry 4.0 environment. 2. Distinguish remote audits, e-QMS auditing, automated evidence collection, process mining, IoT-based auditing, blockchain traceability, and AI-assisted auditing. 3. Evaluate the reliability, integrity, security, privacy, and relevance of digital audit evidence. 4. Apply a risk-based framework for planning and conducting technology-enabled audits. 5. Identify cybersecurity, privacy, bias, explainability, and data-integrity risks. 6. Interpret dashboards, process-mining results, sensor data, and algorithmic outputs without surrendering professional judgement. 7. Recommend governance and control practices for responsible digital auditing. From Electronic Records to Intelligent Audit SystemsThe development of digital auditing can be understood as a progression. In the first stage, organizations replace paper records with electronic documents. In the second, e-QMS platforms integrate procedures, nonconformity records, corrective actions, training, supplier information, and audit schedules. In the third, systems exchange data automatically through enterprise resource planning (ERP), manufacturing execution systems (MES), laboratory information management systems (LIMS), and customer platforms. In the fourth, analytics and AI identify patterns, anomalies, and likely control failures. This progression changes the auditor’s object of examination. A paper-based audit may ask whether an inspection record exists and is signed. A digital audit also asks who could alter the record, whether the time stamp is system-generated, whether edits are logged, whether the inspection device was calibrated, whether the data was transferred completely, and whether the system permits an inspection to be bypassed. Digital evidence should be assessed using the same fundamental principles as other audit evidence: relevance, sufficiency, reliability, authenticity, and traceability. Technology can strengthen evidence by providing time-stamped histories and large-scale analysis, but it can also create false confidence. A large volume of data is not automatically good evidence. If the source is poorly governed, quantity merely amplifies error. Important Concepts and TerminologyA remote audit uses communication and information technologies to assess activities, records, personnel, or locations without the audit team being physically present for all or part of the audit. An e-QMS is an electronic quality management system that manages controlled documents, records, audits, risks, incidents, training, corrective actions, and related workflows. Automated evidence collection is the programmed retrieval of records, system logs, transactions, configurations, or measurements for audit purposes. Process mining analyses event logs to reconstruct how a process actually operates and compare it with an intended process model. The Internet of Things is a network of connected physical devices that sense, transmit, receive, or act on data. An audit trail is a chronological record showing relevant actions, changes, users, and system events. Data integrity means that data remains complete, accurate, consistent, attributable, legible, contemporaneous, original or appropriately preserved, and protected from unauthorized alteration. A digital dashboard presents selected indicators, trends, alerts, and exceptions for monitoring and decision-making. Blockchain traceability uses a distributed ledger to record transactions or events in linked, tamper-evident blocks. Artificial intelligence refers broadly to computational methods that perform tasks associated with human reasoning, prediction, classification, perception, or language processing. Machine learning is an AI approach in which models identify patterns from data rather than being governed entirely by explicit rules. Explainability concerns the extent to which the reasoning or influential factors behind an algorithmic output can be understood. Model drift occurs when the relationship between input data and the outcome being predicted changes over time, reducing model performance. Data lineage describes the origin, transformation, movement, ownership, and use of data throughout its life cycle. Zero trust is a security approach that assumes no user, device, or network connection should be trusted automatically. The Digital Audit ArchitectureA useful way to understand digital auditing is to examine five connected layers. The first layer is the physical and operational layer. It includes machines, instruments, production lines, warehouses, laboratories, employees, and suppliers. Sensors, scanners, cameras, and devices generate observations at this level. The second layer is the data layer. It includes databases, event logs, master data, metadata, time stamps, interfaces, and data-storage arrangements. Auditors must examine whether data is complete, correctly formatted, synchronized, retained, and protected. The third layer is the application layer. ERP, MES, e-QMS, LIMS, maintenance systems, and supplier portals execute workflows and business rules. The auditor examines authorization, validation, configuration, segregation of duties, exception handling, and change control. The fourth layer is the analytics and intelligence layer. Dashboards, statistical models, process-mining tools, predictive-maintenance models, and AI systems convert data into alerts or recommendations. Here, the important questions concern model purpose, training data, performance, bias, explainability, and human review. The fifth layer is the governance layer. It includes policies, accountability, risk management, cybersecurity, privacy, regulatory obligations, competency, records retention, and continual improvement. A technically impressive system remains weak if governance is unclear. These layers are interdependent. A reliable AI alert cannot compensate for defective sensor calibration. An accurate dashboard cannot compensate for unauthorized access. A blockchain cannot establish that the original entry was truthful. Digital audit planning should therefore trace evidence from its origin to the decision made from it. Remote Audits: Capability and BoundariesRemote auditing is valuable when sites are geographically dispersed, travel is restricted, specialists are located elsewhere, or frequent surveillance is needed. It can reduce cost and provide rapid access to centralized records. Screen sharing, secure portals, video observation, electronic interviews, digital sampling, and live demonstrations can support substantial audit coverage. However, remote methods can reduce the auditor’s ability to observe informal practices, physical conditions, non-verbal communication, workarounds, and the relationship between documented procedures and actual behaviour. The auditee may control the camera view, select records in advance, or temporarily improve conditions during a virtual tour. Connectivity problems may also interrupt evidence collection. A sound remote-audit plan should define the scope of remote activities, required technology, identity verification, secure file exchange, communication protocols, contingency arrangements, sampling method, and circumstances requiring physical verification. High-risk activities, safety-critical operations, suspected fraud, unexplained data anomalies, and sensitive physical controls may require an on-site component. Remote auditing is consequently a method, not a lower standard of auditing. The evidence must still support valid conclusions. When a limitation cannot be overcome remotely, it should be recorded rather than concealed in an overly positive audit report. e-QMS Platforms and Automated Evidence CollectionAn e-QMS can integrate the audit cycle. The system may generate an annual programme, assign auditors, issue checklists, link findings to process owners, track corrective actions, verify effectiveness, and provide management reports. Integration reduces duplicate entry and improves visibility of overdue actions. The auditor should not assume that system workflow equals process control. Important examination areas include: - role-based access and segregation of duties;
- approval authority and electronic signatures;
- version control for procedures and forms;
- retention, backup, and restoration;
- change management and validation;
- interface controls between systems;
- exception reports and rejected transactions;
- audit-log completeness and protection;
- controls over master data and reference values.
Automated evidence collection can retrieve all transactions in a defined period rather than relying only on small samples. It can identify missing approvals, duplicate records, unusual timing, expired training, repeated deviations, or corrective actions closed without effectiveness verification. Yet automation creates its own sampling risks. The query may exclude relevant records, filters may be misunderstood, or a data extract may not match the production database. The auditor should preserve the query logic, extraction date, population definition, filters, and reconciliation to source totals. Process Mining and the Discovery of Actual WorkTraditional process audits generally compare interviews and selected records with documented procedures. Process mining adds an empirical view by using event logs containing a case identifier, activity name, and time stamp. For example, a purchase order may be traced from requisition through approval, receipt, inspection, invoice matching, and payment. Process-mining analysis can reveal variant paths, rework loops, bottlenecks, skipped approvals, excessive waiting time, and activities performed out of sequence. Conformance checking compares the observed process with the approved model. Discovery methods construct a process model from the event log when the documented model is incomplete. Interpretation requires caution. An event log may record only system events, not physical actions. Similar activities may have different names in different systems. Missing or incorrect case identifiers can create false paths. A process that appears nonconforming may reflect a justified emergency route that was not represented in the model. Process mining is therefore an aid to inquiry, not a substitute for understanding context. A mature audit uses process mining to select interviews, transactions, and site observations. It asks why a high-risk variant occurs, whether it is authorized, whether it has adverse outcomes, and whether the documented process should be revised. IoT Data, Smart Operations, and Audit EvidenceIoT devices allow organizations to monitor temperature, humidity, pressure, vibration, energy use, location, machine status, and environmental conditions. In regulated or quality-sensitive operations, continuous monitoring can provide more complete evidence than occasional manual readings. Alerts may support preventive action before a specification is breached. Auditors should evaluate the measurement chain. This includes sensor selection, calibration, identification, installation, sampling frequency, time synchronization, transmission reliability, data storage, alarm thresholds, escalation, and response. A sensor reading is not meaningful merely because it appears on a screen. The auditor should establish whether the instrument measures the relevant variable accurately and whether missing or implausible readings are detected. Important questions include: What happens when connectivity fails? Can an operator suppress an alarm? Are alarms acknowledged and investigated? Are devices updated securely? Is data transmitted in a way that prevents unauthorized alteration? Are environmental readings linked to the correct batch, equipment, or location? IoT evidence is particularly useful for trend analysis and continuous auditing, but it should be combined with physical verification and maintenance records. A disconnected sensor can create a period of apparent stability precisely when no information is being collected. Dashboards and the Problem of Compressed InformationDashboards convert large data sets into indicators, charts, traffic lights, and alerts. They support management review and help auditors identify areas requiring attention. Useful quality indicators may include defect rates, first-pass yield, complaint trends, supplier performance, overdue corrective actions, audit finding recurrence, and process capability. The design of a dashboard influences behaviour. If managers reward only low defect numbers, personnel may under-report defects. If a dashboard averages results across plants, serious local variation may disappear. If indicators are updated at different intervals, comparisons may be misleading. An auditor should therefore examine indicator definitions, numerator and denominator, data source, calculation logic, update frequency, thresholds, ownership, and action rules. A dashboard is an interpretation layer, not the evidence itself. Audit work should trace selected indicators to underlying records and determine whether adverse information has been excluded. Visual simplicity must not be confused with analytical validity. Blockchain Traceability: Promise and QualificationBlockchain or distributed-ledger systems can create a shared, time-sequenced record of transactions across organizations. In supply chains, they may support traceability of material movement, certifications, ownership, or custody events. Because subsequent alteration is difficult to conceal, the technology can strengthen confidence in the recorded sequence. Its limitations are equally important. Blockchain generally protects the ledger after data entry; it does not guarantee that the initial data was accurate. A fraudulent or defective input can become a permanently preserved error. Identity management, access permissions, consensus arrangements, interoperability, data privacy, and governance remain essential. Organizations must also determine which information should be stored on-chain and which should remain in controlled external systems. Auditors should examine the control over data entry, identity of participants, validation of transactions, exception handling, synchronization with physical events, and procedures for correcting erroneous records. Blockchain is a traceability mechanism, not a universal replacement for inspection or supplier control. AI-Assisted AuditingAI can support audit planning by ranking processes according to risk, identifying unusual transactions, summarizing records, matching requirements to evidence, predicting overdue corrective actions, and detecting anomalies in images or sensor data. Natural-language tools may help search large document collections and compare revisions. Machine-learning models may identify combinations of conditions associated with defects or complaints. The auditor must distinguish assistance from delegation. An AI system can prioritize a sample, but the auditor remains responsible for determining whether the sample is appropriate. An algorithm can flag an anomaly, but the anomaly may result from a legitimate change in production. A language model can summarize evidence, but it may omit qualifications or generate an inaccurate statement. A responsible AI-assisted audit examines: 1. Purpose and materiality: What decision does the model support, and what is the consequence of error? 2. Data quality and representativeness: Are training and operational data complete, current, relevant, and sufficiently diverse? 3. Validation: Has performance been tested against known cases and independently reviewed? 4. Explainability: Can the auditor understand the principal factors influencing an output? 5. Human oversight: Who can challenge, override, or suspend the model? 6. Monitoring: Are false positives, false negatives, drift, and changing conditions tracked? 7. Security and privacy: Are data, prompts, models, and outputs protected? 8. Recordkeeping: Can the organization reproduce the model version, input data, output, and decision made? Bias may enter through historical data, proxy variables, unequal reporting practices, or labels created by inconsistent human judgements. Explainability is especially important where AI affects supplier approval, employee evaluation, product release, or escalation of nonconformities. An opaque score should not be treated as an unquestionable audit conclusion. A Risk-Based Method for Planning a Digital AuditA practical digital-audit method begins with understanding the process and its intended outcomes. The auditor maps critical controls, information flows, systems, users, interfaces, and external dependencies. Risks are then assessed according to the potential effect on conformity, customers, safety, privacy, continuity, and decision quality. The next step is to classify evidence sources. Direct observation, validated system records, protected audit trails, and corroborated sensor data may provide stronger evidence than uncontrolled spreadsheets or screenshots. The auditor defines data requirements, access rights, extraction procedures, security arrangements, and criteria for corroboration. During fieldwork, automated analytics should be used to identify exceptions and patterns. The auditor investigates rather than automatically converting every exception into a finding. Interviews and observation provide context, while targeted tests confirm whether digital controls operate as designed. At the conclusion, findings should identify the process requirement, objective evidence, nature of the gap, risk, and responsible process. Reports should distinguish system defects from user errors and data-quality problems from process-performance problems. Follow-up should verify not only that a corrective action was completed but also that the underlying digital control is effective and has not created new risks. Cybersecurity, Privacy, and Data GovernanceDigital auditing increases the movement and concentration of sensitive information. Audit files may contain personal data, supplier prices, designs, customer complaints, access credentials, medical information, or commercially confidential records. Secure transfer, least-privilege access, encryption, retention limits, and controlled deletion are therefore professional obligations as well as technical practices. Cybersecurity risks include compromised remote-access tools, phishing, ransomware, insecure application programming interfaces, excessive privileges, vulnerable IoT devices, and manipulation of audit evidence. A quality auditor need not become a penetration tester, but should recognize whether cybersecurity weaknesses undermine the reliability or availability of quality information. Privacy requires purpose limitation, data minimization, lawful processing, appropriate notice, and protection against unnecessary exposure. Audit teams should avoid downloading entire databases when a restricted extract is sufficient. Personal information should be masked or redacted where identity is not relevant to the audit conclusion. Data governance should assign ownership for data definitions, quality, access, retention, lineage, and correction. A clear governance model prevents disputes about which system is authoritative and supports reproducibility of audit results. Illustrative Case: Remote and AI-Assisted Audit of a Medical-Device SupplierBackgroundThis illustrative case concerns Meditron Components, a fictional supplier of precision-moulded parts for a medical-device manufacturer. The supplier uses an e-QMS, IoT temperature monitoring, automated inspection cameras, and a machine-learning model that predicts dimensional defects. A customer schedules a remote audit because travel is restricted and recent complaints indicate intermittent fit problems. Situation and ProblemMeditron’s dashboard shows a declining defect rate and no overdue corrective actions. However, the customer’s complaint data suggests that failures are concentrated in one product family. The remote audit team must determine whether the supplier’s digital evidence is reliable and whether the AI model is masking a developing problem. AnalysisThe auditors review the dashboard definitions and discover that the defect-rate denominator excludes batches placed on hold. The e-QMS automatically closes corrective actions when a replacement record is uploaded, although effectiveness review is not mandatory. Process mining identifies a recurring route in which inspection results are entered after shipment documentation is prepared. The auditors also find that the vision model was trained mainly on data from an earlier mould design. IoT records show several short gaps during a period of high humidity. The system displayed the last valid reading rather than a clear communication-failure warning. Access-log review indicates that supervisors share one account during night shifts, weakening attribution. Approach and SolutionThe audit team does not reject all digital evidence. Instead, it classifies each source according to reliability and corroborates critical information with raw inspection files, equipment records, batch histories, customer complaints, and interviews. The supplier is required to revise dashboard logic, prevent automatic closure without effectiveness approval, create a visible sensor-failure state, retrain and validate the AI model against current product data, and introduce individual user accounts. A targeted physical audit is scheduled for the moulding and inspection areas. OutcomeThe investigation identifies that the apparent quality improvement was partly a reporting artefact. Several batches require additional review, and the supplier updates its control plan. The AI model becomes a screening tool rather than an automatic release decision. Management also establishes a data-governance group involving quality, information security, engineering, and operations. Learning PointsThe case demonstrates that dashboards may conceal denominator changes, automation may close workflow steps without proving effectiveness, and AI performance depends on current and representative data. It also shows why sensor availability, user attribution, and process context are essential to trustworthy digital auditing. Common Failure Modes and Professional JudgementOrganizations often begin with technology rather than a defined control problem. They purchase dashboards without agreeing on indicator definitions, deploy sensors without calibration and maintenance responsibilities, or introduce AI without deciding who owns the decision. Another common mistake is treating system access as evidence of process performance. A record may exist because a field was completed, not because the underlying activity was properly performed. Other failures include excessive reliance on screenshots, inadequate preservation of queries and system versions, untested interfaces, uncontrolled spreadsheets, weak password practices, unreviewed algorithmic recommendations, and failure to investigate repeated exceptions. Auditors may also overreact to novelty, assuming that any digital system is unreliable, or underreact to novelty, assuming that automation is inherently objective. Professional judgement requires proportionality. High-impact automated decisions require stronger validation, monitoring, explainability, and human oversight than low-risk administrative automation. The auditor should document assumptions, limitations, alternative explanations, and the basis for concluding that evidence is sufficient. Governance, Standards, and Emerging DirectionsDigital auditing should be aligned with the organization’s quality-management requirements, information-security controls, privacy obligations, records-management rules, and sector-specific regulations. Quality, IT, cybersecurity, data protection, internal audit, legal, and operational functions should cooperate rather than operate in separate control silos. Emerging developments include continuous auditing, digital twins for process simulation, edge analytics close to the point of data generation, privacy-enhancing computation, synthetic data for testing, computer vision for inspection, and federated learning across organizations. These developments may improve speed and coverage, but they also increase the importance of model governance, system resilience, skills, and accountability. The strongest future capability will not be fully autonomous auditing. It will be augmented auditing in which technology expands the auditor’s reach while experienced professionals interpret evidence, challenge assumptions, investigate causes, and make accountable judgements. Best Practices for Digital AuditingEffective organizations generally: - begin with process and risk, not with a technology purchase;
- maintain a documented inventory of systems, interfaces, devices, models, and data owners;
- validate critical applications and preserve configuration and version histories;
- reconcile extracts to source populations and retain reproducible query logic;
- combine automated analysis with interviews, observation, and physical verification;
- apply least-privilege access, strong authentication, encryption, backups, and secure transfer;
- establish rules for privacy, retention, redaction, and third-party access;
- validate AI models, monitor drift, test for bias, and require human review for material decisions;
- make sensor failures, missing data, and exceptions visible rather than carrying forward stale values;
- train auditors in data literacy, cybersecurity awareness, process analysis, and responsible AI;
- record limitations and avoid overstating conclusions derived from dashboards or algorithms.
Chapter SummaryDigital auditing extends quality auditing into systems, data, devices, algorithms, and digital governance. Remote audits and e-QMS platforms can improve access and workflow control, while automated evidence collection increases coverage when queries and populations are properly controlled. Process mining reveals actual process paths, IoT data supports continuous monitoring, dashboards aid risk-based attention, and blockchain can strengthen transaction traceability. None of these technologies guarantees accurate evidence. Reliable digital auditing depends on data lineage, integrity, access control, cybersecurity, privacy, validation, and corroboration. AI can assist in risk assessment, anomaly detection, document analysis, and predictive activities, but human judgement remains essential. Bias, model drift, opacity, poor training data, and automation bias must be actively managed. The appropriate objective is not technology for its own sake, but trustworthy, explainable, secure, and decision-relevant audit evidence. Key TermsRemote audit: An audit conducted partly or wholly through communication and information technologies. E-QMS: An electronic system for managing quality processes, records, workflows, and controls. Automated evidence collection: Programmed retrieval of records, logs, transactions, or measurements for audit use. Process mining: Analysis of event logs to reconstruct and evaluate actual process execution. IoT: A network of connected physical devices that collect, exchange, or act on data. Audit trail: A chronological record of system activities and changes. Data integrity: The preservation of data accuracy, completeness, consistency, attribution, and protection from improper alteration. Data lineage: The history of data origin, movement, transformation, ownership, and use. Dashboard: A visual display of selected performance indicators, trends, and alerts. Blockchain traceability: Use of a distributed, tamper-evident ledger to record transactions or events. Artificial intelligence: Computational capability used for tasks involving prediction, classification, reasoning, or perception. Machine learning: An approach in which models learn patterns from data. Explainability: The extent to which the basis of an algorithmic output can be understood. Model drift: Decline in model performance caused by changes in data or operating conditions. Cybersecurity: Protection of systems, networks, devices, and information from unauthorized access or disruption. Privacy: Protection of personal information and appropriate control over its collection and use. Data governance: Assignment of accountability for data quality, access, definitions, retention, and use. Zero trust: A security principle requiring verification rather than automatic trust of users or devices. Continuous auditing: Frequent or ongoing audit analysis supported by automated data access and monitoring. Automation bias: The tendency to accept automated outputs uncritically. Review QuestionsShort Answer Questions1. What distinguishes a digital audit from a conventional audit based only on electronic documents? 2. Why should a dashboard not be treated as the audit evidence itself? 3. State two controls required for reliable IoT-based audit evidence. 4. What is process mining, and what type of information does it use? 5. Define model drift in the context of AI-assisted auditing. Long Answer Questions1. Explain the five-layer architecture of a digital audit and show how weaknesses in one layer can affect the reliability of conclusions. 2. Discuss the benefits and limitations of remote audits and identify circumstances requiring physical verification. 3. Examine how e-QMS platforms and automated evidence collection can improve audit effectiveness while creating new control risks. 4. Explain the governance requirements for responsible use of AI in quality auditing. 5. Discuss cybersecurity, privacy, and data-integrity considerations in an Industry 4.0 audit. Analytical / Application Questions1. A factory dashboard reports zero overdue corrective actions, but interviews reveal several unresolved problems. Design an audit procedure to test the reliability of the dashboard and e-QMS workflow. 2. An AI model flags a supplier as high risk, but the supplier disputes the result. Explain how an auditor should assess data quality, bias, explainability, validation, and human review before reaching a conclusion. 3. During a remote audit, IoT temperature data shows stable conditions, but the audit trail contains communication gaps. Analyse the evidence and recommend additional tests and corrective controls. 22. Multiple Choice Questions with AnswersQuestion 1Which statement best describes the central responsibility of an auditor using digital evidence? A. Accepting system-generated information because it is more objective than interviews B. Assessing whether the evidence is relevant, reliable, complete, secure, and traceable C. Replacing all physical observations with dashboard analysis D. Using the largest possible data set regardless of data quality Correct Answer: B. Assessing whether the evidence is relevant, reliable, complete, secure, and traceable Question 2What is a principal limitation of a remote audit? A. It cannot examine any electronic record B. It always requires more travel than an on-site audit C. It may restrict observation of informal practices and physical conditions D. It prevents auditors from interviewing personnel Correct Answer: C. It may restrict observation of informal practices and physical conditions Question 3Which information is essential for reconstructing a process through process mining? A. Only the organization chart B. A case identifier, activity name, and time stamp in event logs C. A blockchain record for every transaction D. A dashboard containing only monthly averages Correct Answer: B. A case identifier, activity name, and time stamp in event logs Question 4Why does blockchain traceability not guarantee that recorded information is correct? A. Blockchain records cannot contain time stamps B. The ledger protects the record after entry but does not validate the truth of the original input C. Blockchain systems automatically delete old records D. Blockchain cannot record supply-chain events Correct Answer: B. The ledger protects the record after entry but does not validate the truth of the original input Question 5An IoT dashboard displays the last valid temperature reading during a network outage. What is the principal audit concern? A. The reading may be interpreted as current when no current measurement exists B. The sensor has necessarily been over-calibrated C. The dashboard contains too many indicators D. The system has become a blockchain Correct Answer: A. The reading may be interpreted as current when no current measurement exists Question 6Which practice most directly supports reproducibility of automated evidence collection? A. Saving only a screenshot of the final chart B. Allowing each auditor to use undocumented filters C. Preserving query logic, extraction date, population definition, and reconciliation to source totals D. Deleting the source data after producing the report Correct Answer: C. Preserving query logic, extraction date, population definition, and reconciliation to source totals Question 7What is automation bias in AI-assisted auditing? A. A model’s ability to identify biased data B. The tendency to accept an automated output without adequate critical evaluation C. The deliberate encryption of audit evidence D. The use of human judgement to challenge an algorithm Correct Answer: B. The tendency to accept an automated output without adequate critical evaluation Question 8Which control is most appropriate when an AI model influences a material quality decision? A. Allowing the model to operate without monitoring once validated B. Prohibiting all use of historical data C. Requiring validation, monitoring, explainability, and accountable human oversight D. Replacing the audit trail with a verbal explanation Correct Answer: C. Requiring validation, monitoring, explainability, and accountable human oversight Question 9During an audit, a process-mining tool identifies an unusual process path. What should the auditor do first? A. Automatically issue a major nonconformity B. Delete the variant as irrelevant C. Investigate the path, validate the event-log quality, and determine whether the variation is justified D. Assume that the documented process model is incorrect Correct Answer: C. Investigate the path, validate the event-log quality, and determine whether the variation is justified Question 10Which principle best reflects responsible digital-audit planning? A. Begin with technology selection and then identify possible uses B. Begin with process risks, intended decisions, evidence requirements, and control objectives C. Use AI outputs as final audit conclusions D. Collect all available personal data to maximize audit coverage Correct Answer: B. Begin with process risks, intended decisions, evidence requirements, and control objectives
|