IBM A1000-118 Practice Exam 2025: Latest Questions

A project team is explaining AI to business stakeholders. Which statement best distinguishes AI from traditional rule-based automation?

A healthcare team wants to group patient records into segments to tailor outreach programs, but they do not have labeled outcomes. Which machine learning approach fits best?

A team wants to build a chatbot that answers questions using content from internal policy documents. Which IBM Watson capability is most directly aligned to this need?

A retailer uses an AI model to recommend products. Which action best supports responsible AI by improving transparency to users?

A supervised learning model performs very well on training data but significantly worse on new data. What is the most likely issue and best immediate mitigation?

A data science team is building a binary classifier for fraud detection where only 1% of transactions are fraudulent. Which metric is most informative for evaluating the model’s ability to catch fraud while controlling false alarms?

An organization wants to deploy an AI assistant that answers employee questions. The assistant must use approved internal documents and should reduce hallucinations by citing sources. Which architecture choice best supports this requirement?

A company notices that a sentiment analysis model’s performance gradually degrades after a marketing campaign changes the language customers use in reviews. What is the best next step?

A bank uses an AI model to recommend loan approvals. An audit finds that applicants from a protected group are approved at a significantly lower rate, even when controlling for income and credit history. Which approach best addresses this issue while aligning with responsible AI practices?

A team builds a model and reports excellent validation results. Later, production performance is far worse. Investigation reveals that a feature used during training was derived from information only available after the prediction time (e.g., a post-event status). What is the root cause?

A team is explaining AI concepts to non-technical stakeholders. Which statement best differentiates AI, machine learning (ML), and deep learning (DL)?

A product owner asks why a classification model sometimes produces probabilistic outputs (e.g., 0.72) instead of a clear label. What is the best explanation?

A business team wants a chatbot that can answer questions from internal policy documents with minimal custom ML training. Which approach is most appropriate?

A dataset for loan approval contains 95% 'approved' and 5% 'denied'. A model achieves 95% accuracy by always predicting 'approved'. Which metric is most appropriate to evaluate performance on the minority class?

A team is building a computer vision model and notices the training loss keeps decreasing while validation loss starts increasing after several epochs. What is the most likely issue and best next step?

A healthcare organization wants to reduce the risk of exposing personal data when building an AI solution. Which practice best supports privacy-by-design?

A company deploys an AI model for fraud detection and wants to ensure it remains reliable over time. Which monitoring approach is most appropriate after deployment?

A team is designing an enterprise Q&A assistant. Users must be able to trace answers back to source passages for audit purposes. Which design decision best supports this requirement?

A hiring-screening model shows significantly different false-negative rates across two demographic groups. The business goal is to reduce disparate impact while keeping performance acceptable. Which action is the most appropriate first step?

An organization wants to use a foundation model to summarize confidential internal reports. Security requires that the model does not retain or leak proprietary content and that access is controlled. Which architecture choice best meets these needs?

A retail team wants to quickly group customers into segments based on purchasing behavior without having pre-labeled outcomes. Which machine learning approach best fits this goal?

A team built a sentiment model that performs well on English product reviews but fails on reviews that include common slang and misspellings. Which improvement is the most appropriate first step?

A bank wants an AI-assisted customer support solution that can answer FAQs and also pass complex cases to a live agent with conversation context preserved. Which architecture is most appropriate using IBM Watson technologies?

An HR screening model is found to disproportionately reject qualified candidates from a protected group. The business requires a mitigation that improves fairness while still keeping the model usable. Which action is the best practice?

A team is building an AI workflow and wants reproducibility: the ability to trace which data, features, and model configuration produced a given prediction used in a business decision. Which practice best supports this requirement?