IBM Assessment: Foundations of AI Practice Exam 2025: Latest Questions

A retailer wants to add a chatbot to answer common customer questions (order status, return policy, store hours) and route complex issues to a human agent. Which IBM Watson service best fits this requirement?

Which statement best describes the difference between artificial intelligence (AI) and machine learning (ML)?

A team is building a binary classifier (approve/deny) and observes 95% accuracy. However, only 5% of applications are actually fraudulent, and the model misses most fraud cases. Which metric is most useful to evaluate performance on the minority (fraud) class?

An organization wants to reduce the risk of bias in an AI-driven hiring screen. Which action is the best example of an ethical best practice?

A company has thousands of PDFs (policies, manuals, FAQs) and wants an employee search tool that returns relevant passages and answers with citations. Which IBM Watson capability is the most appropriate starting point?

A data scientist is building a model to predict house prices. Which approach is most appropriate for this problem type?

A model performs very well on training data but significantly worse on new, unseen data. Which issue is most likely occurring?

A team wants to improve trust in an AI system by helping business users understand which factors most influenced individual predictions. Which technique best supports this goal?

You are designing an enterprise virtual agent. Requirements: (1) conversational interface, (2) answers grounded in internal documents, and (3) ability to escalate to a human agent. Which architecture best meets these needs using IBM Watson services?

A credit-risk model shows strong performance in development, but after deployment, the approval rate for a certain region drops sharply while input distributions (income, employment length) shift compared to training data. What is the most likely cause and best next step?

A project team is explaining AI to non-technical stakeholders. Which description best distinguishes Artificial Intelligence (AI) from Machine Learning (ML)?

A retail chatbot must answer questions like return policy, order status, and store hours. The team wants an IBM Watson service designed for conversational experiences. Which service is the best fit?

A data scientist evaluates a binary classifier and observes high accuracy, but the dataset is 98% negative and 2% positive. Which metric is typically more informative for assessing performance on the minority (positive) class?

An insurance company wants to automatically extract key fields (e.g., claimant name, policy number, incident date) from scanned claim forms. What is the most appropriate high-level AI approach?

A team built a document Q&A capability for internal HR policies. They need to search across policy PDFs and return relevant passages as answers with citations. Which IBM Watson service is most aligned with this requirement?

A model performs very well on training data but significantly worse on validation data. Which statement best describes what is happening and a common mitigation?

A credit decisioning model shows a higher false-negative rate for one demographic group. The business wants to address potential unfairness. Which action is a best practice?

A call center wants to analyze customer phone calls to identify common complaint themes. They plan to convert audio to text and then extract topics and sentiment from transcripts. Which combination best matches the needed capabilities?

A hospital is considering an AI tool to recommend treatments. Clinicians require reasons behind recommendations and the ability to audit inputs that influenced each output. Which model choice and practice best aligns with this requirement?

A team built a churn prediction model. Validation performance is excellent, but production performance quickly degrades because customer behavior changes over time. What is the best next step to address this issue?

A customer support team wants to quickly categorize incoming emails into predefined issue types (e.g., billing, technical problem, account access) using an IBM Watson service. Which approach is most appropriate?

A team is building a model to predict equipment failure. Failures are rare (about 1% of records). Which evaluation metric is generally MOST informative to understand how well the model identifies failures?

A bank is deploying an AI system to assist with credit decisions. Which practice BEST supports the principle of transparency for affected customers?

A data scientist sees excellent training performance but poor validation performance after adding many features and increasing model complexity. What is the MOST likely issue and the best next step?

An enterprise wants to build a retrieval-augmented generation (RAG) solution that answers employee questions using internal policy documents. Which architecture choice MOST helps reduce the risk of the model generating answers not grounded in the company’s content?