A hospital's clinical informatics team has been reviewing an AI-assisted diagnostic application with attending physicians. The review reveals that the model's false-positive rate in flagging abnormal scans is significantly above the acceptable clinical threshold.Before making the go/no-go decision for deployment, what should the team do next?
Answer(s): B
In PMI-CPMAI's model evaluation framework, performance results must always be interpreted in the context of business and clinical objectives. A high false-positive rate has direct implications for clinical workflow, patient anxiety, and resource consumption. Before deciding to tune or retrain, the team and physicians must first reevaluate whether the observed error rate is acceptable within the intended use case—for example, determining whether the system should serve as a decision-support tool or a primary screener. Technical improvements come after confirming the required performance level and error trade-offs.
A metropolitan transit authority plans to deploy an AI-driven scheduling and service improvement platform. The project manager must develop a business case to secure capital funding approval from the board.Which method will produce the most compelling evidence to support the funding request?
Answer(s): D
PMI-CPMAI business case guidance emphasizes that funding requests are strengthened by concrete, evidence-based results rather than theoretical projections. A pilot program allows the team to implement the AI system at limited scale, collect operational metrics such as response time reduction and throughput improvement, and compare pre- and post-AI KPIs. This provides measurable, context-specific evidence that directly supports funding justification. Case studies and workshops are indirect and qualitative; ROI projections alone remain hypothetical without empirical support from a real implementation.
An international healthcare provider is building an AI model to improve diagnostic accuracy by integrating data from partner hospitals across multiple jurisdictions, each with its own data protection laws.Which step is most critical to ensure compliant and ethical deployment?
PMI-CPMAI places Privacy Impact Assessments (PIAs) at the center of responsible AI governance when AI systems process sensitive personal data across multiple legal contexts. A PIA systematically identifies how personal data is collected, processed, stored, shared, and deleted; maps applicable regulations; evaluates privacy risks; and guides the design of safeguards such as data minimization, anonymization, and access controls. Without a properly conducted PIA, the project risks legal non-compliance and loss of stakeholder trust, regardless of model accuracy or financial prudence.
An autonomous fleet management AI that previously performed well has recently seen a notable uptick in prediction errors. The model code has not been modified since the last deployment.What is the most appropriate first diagnostic step?
PMI-CPMAI monitoring and performance management guidance establishes that when a deployed model shows unexpected degradation without code changes, data drift or concept drift is the primary hypothesis. Teams should compare current production data distributions against training and validation baselines to detect feature shifts, population changes, or evolving patterns. Once drift is confirmed, the team can make informed decisions about retraining, recalibration, or architectural adjustment. Preprocessing checks are useful when suspecting a technical bug, but the question describes gradual, real-world degradation—the hallmark of drift.
A defense contractor is integrating AI into a predictive parts-ordering system for aircraft maintenance. The project manager identifies a risk that a sole-source sensor supplier could delay the project significantly.What initial step should the project manager take to reduce this risk?
PMI-CPMAI risk and dependency management guidance emphasizes that when AI projects depend on specialized external inputs—such as proprietary sensors, cloud services, or third-party data streams—project managers should proactively diversify suppliers to reduce single-source vulnerability. Identifying and engaging multiple suppliers early creates redundancy, reduces procurement-related schedule risk, and improves resilience. Increasing internal resources does not address the external supplier risk. Contingency planning is important but represents a reactive response; risk reduction through diversification is the preferred first step.
A financial services company is evaluating the feasibility of an AI-powered chatbot to handle retail customer inquiries. The project team is focused on scalability and quality of customer experience.What presents the highest level of risk to the company in this scenario?
PMI-CPMAI frames data privacy violations as a high-order, potentially non-recoverable risk. For a financial services chatbot processing customer data, breaches can trigger regulatory investigations, heavy fines, and long-term brand erosion—harms that are far harder to remediate than operational challenges such as scalability gaps or integration issues. The guidance stresses embedding privacy-by-design and compliance checks early in the solution lifecycle. Scalability and integration risks are remediable through technical improvements; a privacy breach can result in forced system shutdown and lasting reputational damage.
A precision parts manufacturer is preparing data for an AI defect detection model. The team has discovered that a significant portion of quality records contain errors caused by operators manually entering inspection results.Which approach most effectively prevents these errors from propagating into the training dataset?
Answer(s): A,B
PMI-CPMAI AI data governance guidance identifies controls at the point of data capture as the most effective mechanism for preserving data integrity. Real-time validation rules—enforcing allowed ranges, formats, mandatory fields, and cross-field consistency checks—prevent bad data from entering the system in the first place. Automating data entry processes also reduces the opportunity for manual errors. Periodic audits are reactive and do not prevent errors from reaching training pipelines. Using ML to detect and correct errors adds complexity and relies on the quality of the very data it is attempting to fix.
An investment bank is training an AI model to flag potentially fraudulent trades. The data science team needs to confirm that trade records accurately represent actual transactions before model training begins.What is the most effective method to ensure data accuracy?
Answer(s): A
PMI-CPMAI data governance guidance highlights data lineage as the primary mechanism for establishing data accuracy and trustworthiness in regulated environments. Lineage tools enable teams to trace data back to its source, review each processing and transformation step, and verify compliance with business and regulatory rules. In fraud detection, this traceability is essential for model validation and audit readiness. Federated databases address access architecture, not accuracy. Blockchain and batch cleansing may play supporting roles but do not provide the end-to-end visibility that lineage tools offer.
Post your Comments and Discuss PMI PMI-CPMAI exam prep with other Community members:
Our website is free, but we have to fight against AI bots and content theft. We're sorry for the inconvenience caused by these security measures. You can access the rest of the PMI-CPMAI content, but please register or login to continue.