50 IBM A1000-119 Practice Questions: Question Bank 2025

A retail team wants to quickly categorize customer feedback emails into topics (e.g., shipping, returns, product quality) without building a complex model from scratch. Which approach is most appropriate?

A project team is discussing whether a proposed solution uses AI. The system applies fixed business rules: if a customer spends over a threshold, assign them to a loyalty tier; otherwise assign a lower tier. No model is trained. Which statement is accurate?

A bank wants to reduce the risk of unfair loan decisions and improve transparency for applicants. Which action best supports responsible AI at a foundational level?

A customer support chatbot is designed to answer common questions. Which capability most strongly indicates the chatbot uses Natural Language Processing (NLP)?

A data scientist trains a classification model and observes high accuracy on the training data but noticeably lower accuracy on new, unseen data. What is the most likely issue?

A team needs to build a model to predict house prices based on features such as square footage, location, and number of bedrooms. Which type of machine learning problem is this?

A healthcare organization wants an AI system to assist clinicians by highlighting potentially relevant sections in long patient notes, but the final decision must remain with a doctor. Which design best aligns with this goal?

A product team wants to detect fraudulent transactions in near real time. Fraud cases are rare compared to legitimate transactions. Which metric is most informative to evaluate the model beyond overall accuracy?

A team is building an AI solution and wants to reduce the risk of training-serving skew (the model behaves differently in production than during training). Which practice best helps prevent this issue?

A company wants to use a pre-trained large language model (LLM) to draft responses to customer inquiries. They must minimize the risk of the model generating incorrect policy details. Which approach is most appropriate?

A project team is deciding whether a planned system is AI, traditional software, or both. The system will follow fixed, hand-coded rules for eligibility checks, but will use a model to flag potentially fraudulent applications for review. How should the team classify the overall solution?

A team is building a model to predict whether a customer will churn (yes/no). Which type of machine learning problem is this?

A retail company wants to quickly identify which customer support emails should be routed to billing, shipping, or product support. What is the most appropriate AI approach?

A model shows excellent performance in training but significantly worse performance on new, unseen data. Which issue is most likely occurring?

A healthcare organization wants an AI system to summarize long clinical notes. Clinicians require that summaries avoid fabricating details not present in the note. Which control best helps reduce this risk in a generative AI implementation?

A bank is using an ML model for credit decisions. Regulators require that the bank can explain key factors influencing an individual decision. Which approach best meets this requirement?

A team notices that their model’s performance drops after deployment because customer behavior changes over time. What practice best addresses this issue?

A data scientist splits a dataset into training and test sets. Later, they compute normalization parameters (mean and standard deviation) using the entire dataset and apply them to both sets. What is the primary risk of this approach?

A company is building an AI assistant that answers questions using internal policy documents. Some policies change frequently, and the company needs responses to reflect the latest approved content without retraining the model each time. Which architecture is most appropriate?

A team is evaluating fairness for a hiring-screening classifier. They find different selection rates across demographic groups. Leadership asks for a metric that directly compares selection rates between groups to identify potential disparate impact. Which metric best fits this request?

A team is new to AI and is deciding between supervised and unsupervised learning. They have historical customer records where each record is labeled "churned" or "not churned." What is the most appropriate learning approach?

A product owner asks why model accuracy alone is not sufficient to decide whether a binary classifier is ready for production. Which explanation is most accurate?

A help desk wants to reduce average handling time by automatically routing incoming tickets to the correct category (e.g., password reset, network issue, hardware). What is the most appropriate AI application?

A data scientist trains a model with 99% training accuracy but only 72% validation accuracy. Which issue is most likely, and what is the best first mitigation?

A bank is deploying an AI model to assist loan approvals. Regulators require that adverse action notices include understandable reasons for denial. Which approach best addresses this requirement?

A retailer wants to recommend products to users on its website. The retailer has purchase histories and browsing events, but many items have few interactions (cold-start). Which design is most robust?

A team is preparing data for a supervised learning project and suspects that a single feature ("account_status") might leak the label because it is set after the outcome occurs. What is the best practice to validate and prevent leakage?

An organization deploys a model that performs well initially but gradually produces worse predictions as customer behavior changes. Which operational capability best addresses this problem?

A healthcare provider uses an AI triage model. A review finds it systematically under-prioritizes a minority group due to historical under-treatment in training data. Which governance action is the best first step to reduce harm while maintaining a controlled process?

A company is designing an AI solution to summarize internal documents. Some documents contain sensitive personal data. Which architecture decision best supports privacy and compliance while still enabling model use?

A product manager wants to explain why a chatbot sometimes gives different answers to the same question. Which concept best explains this behavior?

A retail team wants to predict the probability that a customer will churn within 30 days (yes/no). Which machine learning problem type best fits this goal?

A team is planning an AI pilot and wants to choose a metric that aligns with a helpdesk chatbot goal: reduce the number of tickets handled by agents while maintaining customer satisfaction. Which metric is the BEST primary KPI for the pilot?

A company is deploying an AI model that uses customer data. Which practice BEST supports privacy by design?

A data scientist finds a model has 96% accuracy on a dataset where 95% of cases are non-fraud and 5% are fraud. Which evaluation approach is MOST appropriate to understand fraud detection performance?

A team wants to build an AI assistant that answers questions using an internal policy manual that changes weekly. They want to reduce hallucinations and ensure answers reference the latest policy. Which architecture is MOST suitable?

During model training, a team notices training loss continues to decrease, but validation loss starts increasing after several epochs. What is the MOST likely issue and a common mitigation?

An organization wants to operationalize AI governance for multiple teams. Which combination of controls BEST supports responsible AI in production?

A bank builds a loan approval model and later discovers it rejects applicants from a protected group at a higher rate, even though that group has similar repayment outcomes. Which fairness issue is MOST directly indicated?

A generative AI assistant is used by employees to summarize confidential client documents. The security team is concerned about unintentional data exposure through prompts and outputs. Which implementation approach MOST effectively reduces this risk while keeping the solution usable?

A team is building an AI-powered FAQ assistant. They want the system to answer using only the company’s policy documents and to reduce the chance of fabricated answers. Which approach best fits this requirement?

A product owner asks why a binary classifier can have high accuracy but still be a poor solution. The dataset has 95% negatives and 5% positives. Which metric is most appropriate to highlight the model’s ability to detect the positive class?

A stakeholder is confused about what "inference" means in an AI lifecycle. Which statement best describes inference?

A bank is developing a loan approval model and must provide customers with understandable reasons for adverse decisions. Which technique most directly supports this requirement for model transparency?

A model performs well in testing but degrades after deployment because customer behavior changes over time. What is the most likely issue, and what is the best operational response?

A team is training a model for medical image classification. They have limited labeled data but can apply domain-preserving transformations to existing images. Which approach is most appropriate to improve generalization?

A retail company wants to predict next week’s demand per store using features such as promotions, holidays, and past sales. Which model type is most appropriate for the target variable?

A team is evaluating a generative AI feature that summarizes internal meeting notes. Which control best reduces the risk of exposing confidential information in generated summaries shared externally?

A company is building a customer-support chatbot that must cite the exact source passages used to answer each question for audit purposes. Which design choice most directly enables this requirement?

An organization wants to implement governance for ML models used in underwriting. They need strong traceability from a deployed model back to training data, features, evaluation results, and approvals. Which practice best satisfies this requirement?