IBM A1000-077 - Assessment: Foundations of AI Practice Exam 2025: Latest Questions

A team is building an AI system to route customer support emails into categories (billing, technical issue, account access). Which problem type best describes this task?

Which statement best differentiates artificial intelligence (AI) from machine learning (ML)?

A product owner asks why a model performs well on training data but poorly on new, unseen data. What is the most likely issue?

A bank wants to detect and mitigate unintended discrimination in a credit decision model and communicate model behavior to stakeholders. Which IBM capability is most aligned with this need?

A retail company has purchase history labeled as 'churned' or 'not churned'. They want to predict whether a current customer is likely to churn. Which evaluation metric is often most informative when the 'churned' class is rare?

A team wants to build a generative AI application that answers questions grounded in the company’s internal policy documents and reduces hallucinations. Which architecture pattern best fits this requirement?

A chatbot is answering customer questions, but users report it sometimes returns sensitive personal details from past conversations. Which best practice most directly helps prevent this outcome?

A development team needs a managed IBM service to build a conversational interface that can integrate with backend systems and handle intents/entities. Which service is the best fit?

A model is trained on historical hiring data that underrepresents certain groups. Even with high accuracy, the model recommends fewer candidates from those groups. Which technique most directly targets this issue before or during training?

After deploying a customer sentiment model, the team observes performance degradation over time. Investigation shows users started using new slang and abbreviations not present in the training set. What is the most likely cause and recommended response?

A team is deciding whether a new analytics feature should be labeled as AI. The feature only uses fixed, human-written business rules (no learning from data) to classify transactions. How should this capability be described?

A product manager wants a chatbot to answer customer questions using approved policy documents and to cite passages from those documents. What approach best fits this requirement?

A team is evaluating why their model performs much better on training data than on new, unseen data. Which issue most likely explains this behavior?

A bank is building a credit-risk model. Regulators require a clear explanation of why an applicant was declined, including which factors contributed most. Which model choice is generally the best starting point to satisfy this requirement while maintaining reasonable accuracy?

A data science team notices their deployed model’s accuracy is dropping over several months, even though the code and training process have not changed. New customer behavior patterns have emerged. What is the most likely cause?

A retail company wants to automate understanding of customer emails: extract key entities (order number, product name) and determine the request type (refund, shipping status, complaint). Which IBM Watson capability is the most appropriate fit?

A call center wants to transcribe live phone calls and detect customer sentiment in near real time so supervisors can intervene. Which architecture best satisfies the requirement using IBM AI services?

An insurance company plans to use an AI model to recommend claim approvals. Which governance practice is most appropriate to reduce operational and compliance risk before deploying to production?

A healthcare provider wants to use a large language model to draft discharge instructions. The draft must not expose sensitive patient details to unauthorized systems and must minimize the risk of generating unsafe medical advice. Which combination of actions is most appropriate?

A team is building an AI system to screen job applicants. They find that the model’s positive rate differs significantly between protected groups, even when qualifications appear similar. Which metric best quantifies this type of fairness concern at a high level?

A contact center wants to automatically identify whether each inbound call is mainly about billing, technical support, or cancellations. They have thousands of historical call recordings with the final disposition label for each call. Which AI approach best fits this requirement?

A bank is building a loan-approval model. During testing, the model’s overall accuracy is high, but it incorrectly approves some high-risk applicants. The business states that false negatives (rejecting good applicants) are less costly than false positives (approving risky applicants). Which evaluation focus is most appropriate?

A team wants to build a chatbot that answers employee questions by retrieving relevant passages from internal policy documents and generating a response with citations. They want to minimize hallucinations and keep answers grounded in company content. Which architecture pattern best addresses this?

After deploying a churn prediction model, a telecom notices performance degradation over three months. Investigation shows that pricing plans and customer behavior patterns changed significantly due to a new competitor. What is the most likely cause of the degradation?

A healthcare organization is developing an AI model to help prioritize patients for follow-up care. The model may affect access to services. Which governance practice is MOST appropriate to help ensure the model is fair and accountable before deployment?