50 IBM A1000-076 - Assessment: Foundations of AI Practice Questions: Question Bank 2025

A team is explaining why a rules-based chatbot is not considered "AI" in the same way as a machine learning model. Which statement best describes the difference?

A retail analyst wants to predict whether a customer will churn (yes/no) based on past purchases and support interactions. What type of machine learning problem is this?

A developer needs to extract entities such as people, organizations, and locations from unstructured text in customer emails using an IBM Watson service. Which service best fits this requirement?

A bank is evaluating an AI model for loan approvals. Which practice best supports the principle of transparency?

A data scientist notices a model performs very well on training data but poorly on new, unseen data. Which issue is most likely occurring?

A team is building an image classifier. They have limited labeled images but access to a pretrained convolutional neural network. What is the recommended approach to improve performance with limited data?

A product team wants a chatbot that can hand off to a human agent when the user requests account changes and the bot is not confident. Which design best supports this requirement using IBM Watson capabilities?

A team finds that their training dataset under-represents one demographic group, leading to lower accuracy for that group. Which mitigation is most appropriate during model development?

A company deploys an AI model to predict equipment failures. After several months, real-world accuracy declines because operating conditions changed (new machinery and maintenance schedule). What is the most likely cause and best response?

A healthcare organization wants to use a third-party generative AI model to draft patient visit summaries. They must minimize the risk of exposing protected health information (PHI). Which architecture choice is the BEST starting point?

A project team is explaining AI to business stakeholders. Which statement best describes the difference between Artificial Intelligence (AI) and Machine Learning (ML)?

A data scientist evaluates a binary classifier for medical screening where missing a positive case is very costly. Which metric should be prioritized to reduce missed positive cases?

A customer support team wants to automatically extract customer sentiment and key entities (like product names) from incoming emails. Which IBM Watson service capability is the best fit?

A bank trained a loan approval model and notices it performs well on training data but poorly on new applicants. Which issue is most likely occurring?

A retail company wants to build a model in Watson Studio using AutoAI. They have a labeled dataset where the target column is "churn" (yes/no). What is the correct approach to start the AutoAI experiment?

A data pipeline includes customer names and emails. The team wants to train a model while minimizing privacy risk and complying with data protection expectations. Which practice is most appropriate?

A team is choosing between supervised and unsupervised learning for a new initiative. They have thousands of product images with no labels and want to group similar images to discover natural categories. Which approach is most suitable?

A company uses a pretrained deep learning model for image recognition. In production, the camera hardware changes, and the model’s accuracy drops significantly. What is the most likely cause?

A team built a credit risk model and wants to provide an explanation for individual decisions to meet governance expectations. Which approach best supports explainability for a specific prediction?

A developer integrates Watson Assistant for a support chatbot. Users report that the bot often gives irrelevant answers when they ask questions outside the bot’s configured topics. What is the best improvement?

A retail analytics team has 10,000 customer records with a target label indicating whether the customer churned. Only about 3% of customers churned. Which evaluation approach is MOST appropriate to understand model performance on the minority class?

A team wants an AI system that can answer employee questions using internal policy documents and produce citations to the passages used. They do not want to retrain a foundation model. Which approach BEST fits this requirement?

A bank is deploying an AI model to assist with loan approvals. Regulators require the bank to demonstrate why a specific applicant was denied and to identify potentially biased drivers. Which practice MOST directly supports this requirement?

A developer is building a chatbot that must identify intents like "reset password" and extract entities such as account type and urgency from user messages. Which IBM Watson capability is MOST aligned with this need?

A data scientist notices that a model performs extremely well on training data but significantly worse on unseen validation data. What is the MOST likely issue and the BEST first corrective action?

A team is building a document search experience for a help desk. Users want to ask natural-language questions and get the most relevant passages from PDFs and knowledge articles. Which IBM Watson service is MOST appropriate?

An insurance company wants to reduce risk when deploying an AI claims triage model. They plan to continuously collect new claim data and retrain periodically. Which governance practice BEST helps detect performance degradation over time and triggers retraining appropriately?

A team builds a supervised model to predict whether an email is phishing. During testing, they discover that the model used a feature that unintentionally encodes the label, such as a post-incident tag added after investigation. Which problem does this illustrate?

A healthcare organization is selecting an AI approach for analyzing radiology images to detect anomalies. Which model type is MOST appropriate for this kind of image pattern recognition task?

A company plans to deploy a generative AI assistant for customer support. They want to reduce the risk of the assistant producing disallowed content and ensure responses follow brand and compliance guidelines. Which combination is the BEST mitigation approach?

A retail team wants to group customers into segments based on purchasing behavior, but they do not have labels such as "high value" or "low value." Which type of machine learning best fits this goal?

A team built a model that predicts whether an email is spam. Which metric is most directly used to measure the percentage of predicted spam emails that are actually spam?

A company is concerned about bias in a hiring-screening model. Which action best supports transparency and accountability for stakeholders reviewing the system?

Which statement best describes the purpose of a confusion matrix for a binary classification model?

A Watson Assistant bot frequently responds with the wrong answer when users phrase the same question in different ways. What is the most likely improvement to make first?

A team is building a call-center analytics solution. They need to convert audio recordings into text and then analyze customer sentiment. Which combination of IBM Watson services best fits this requirement?

A hospital wants to deploy an ML model to help prioritize radiology cases. They are required to minimize patient data exposure while still enabling model improvement over time. Which practice best aligns with privacy-by-design?

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

An organization trained a model using historical loan approvals. After deployment, it is discovered that applicants from a protected group are being approved at a much lower rate, even when they have similar financial profiles. Which evaluation approach is MOST appropriate to identify and quantify this issue?

A team plans to deploy a conversational AI that must answer questions using content from thousands of internal policy documents and must return citations to the source passages. Which architecture is most appropriate using IBM Watson services?

A product team is deciding whether a new feature should be implemented using AI or traditional software rules. The feature is to automatically categorize customer emails that vary widely in wording. Which statement best supports using AI?

A developer evaluates a binary classifier on an imbalanced dataset where only 2% of cases are positive (fraud). The model achieves 98% accuracy by predicting all transactions as non-fraud. Which metric would be most appropriate to assess model usefulness for detecting fraud?

An HR chatbot should ask clarifying questions when it is uncertain, rather than confidently giving a wrong answer. Which design choice best supports this requirement?

A team needs an IBM Watson service to extract structured information (for example, parties, locations, dates) from unstructured text in insurance claims and then index the results for search. Which Watson capability best matches the extraction requirement?

A data scientist trains a model in Watson Studio and wants to deploy it so multiple applications can call it consistently through a managed endpoint with versioning and governance. Which IBM capability is the best fit?

A team notices their sentiment model performs well during testing but steadily degrades after deployment because customer language and product names change over time. What is the most likely cause and the best response?

A bank is building a credit approval model and must support auditability. Regulators require a clear explanation for each decision, including key factors that influenced the outcome. Which approach best aligns with this requirement?

A team trains a multi-class image classifier and sees excellent training accuracy but much lower validation accuracy. Which action is a best-practice first step to reduce the generalization gap?

A retail company wants to ensure their AI-driven job screening tool is fair across demographic groups. After evaluation, the selection rate for one group is significantly lower than others. What is the most appropriate next step?

An enterprise wants to build an end-to-end solution: ingest documents, extract entities and passages, and provide an internal Q&A search experience. The architecture must also restrict results based on the employee’s role (access control). Which design is MOST appropriate using IBM Watson services?