Technology Project Management Flashcards
7 cards from real CAIC practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 7 Technology Project Management flashcards as text
Which technique helps AI project teams estimate story points for data-dependent tasks with high uncertainty?
Answer: Planning poker with explicit uncertainty ranges
Planning poker surfaces team disagreements and allows uncertainty ranges to be encoded in estimates for AI's inherent unknowns.
A company is deploying an AI model that will affect hiring decisions. The project manager must ensure the team conducts:
Answer: A bias and fairness audit across protected demographic groups
Hiring AI is subject to employment law and ethical standards requiring bias audits across protected groups before deployment.
In MLOps, a CI/CD pipeline for machine learning differs from traditional software CI/CD primarily because it must also:
Answer: Retrain, evaluate, and validate models as part of the pipeline
ML CI/CD must include automated model retraining and evaluation gates, not just code testing and deployment.
Which stakeholder role is MOST critical to define clearly at the start of an AI consulting engagement to prevent scope creep?
Answer: The product owner who controls the backlog and success criteria
A clear product owner with authority over scope and success criteria prevents stakeholders from continuously expanding requirements.
A project team is selecting between building a custom AI model versus using a pre-trained foundation model API. The PRIMARY project management factor favoring the API approach is:
Answer: Faster time-to-value and lower initial development cost
Foundation model APIs reduce development cycles and upfront cost, accelerating delivery even though they introduce vendor dependency.
Which risk category is unique to AI projects compared to traditional software projects?
Answer: Training data quality and availability risk
AI projects uniquely depend on data quality, volume, and access, which can derail the entire project if not secured early.
When a project retrospective reveals that data labeling took 60% longer than estimated, the corrective action for future AI projects should include:
Answer: Building labeling time buffers and exploring semi-supervised or active learning approaches
Adding buffers and exploring label-efficient methods addresses the root cause of data labeling underestimation systematically.