50 IBM A1000-078 - Assessment: Foundations of AI Practice Questions: Question Bank 2025

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

A retail company wants to predict next month’s sales revenue using historical sales data labeled with the actual revenue. Which type of machine learning problem is this?

A business user wants to quickly extract entities and keywords from customer emails without building a custom NLP model. Which IBM Watson service capability is the best fit?

A bank is reviewing potential AI use cases. Which use case is the best example of using AI to improve operational efficiency rather than creating a new product?

A model shows 96% accuracy in detecting fraudulent transactions, but fraud cases are only 1% of all transactions. Which additional metric is most important to review to understand real-world performance?

A data scientist trained a model and achieved excellent results on the training set, but performance drops significantly on new, unseen data. What is the most likely issue?

A team is building an FAQ chatbot for a public website and wants to allow users to ask questions in natural language and get answers from a curated knowledge base. Which approach is most appropriate in IBM Watson Assistant?

A healthcare organization is designing an AI solution to assist clinicians. Which design choice best supports trustworthy AI principles?

A team is training a model to screen job applicants. They notice the model disproportionately rejects candidates from a protected group. What is the BEST next step to address this responsibly?

An organization wants to deploy a text classification model and ensure consistent, repeatable model lifecycle management (tracking data, experiments, and versions) across teams. Which practice best supports this goal?

A business stakeholder asks why an AI assistant sometimes provides incorrect answers even though it was trained on accurate internal documents. Which concept best explains this behavior?

A team wants a chatbot to answer questions strictly using approved policy documents and to cite where the answer came from. Which approach best meets this requirement?

An HR team is building a model to screen resumes. Which action best supports ethical AI and helps reduce unintended bias?

A data scientist trains a classifier and sees 98% accuracy, but the business complains it rarely catches the important positive cases (which are rare). Which metric should the team prioritize to better reflect this requirement?

A team wants to quickly build and deploy a supervised machine learning model on IBM Cloud with minimal code and a guided UI experience. Which IBM service is the best fit?

A customer support team has thousands of PDFs and wants to search them and extract answers to natural language questions without manually building a knowledge base. Which IBM Watson service is most appropriate?

A sentiment analysis model performs well in testing, but after deployment its performance steadily degrades because customer language changes over time (new slang, product names). What is the best practice to address this?

A developer integrates multiple Watson services and wants to securely manage service credentials and rotate them without hardcoding secrets in application code. What is the recommended approach on IBM Cloud?

A bank wants to deploy an AI model for credit decisions. Regulations require that the bank can provide human-understandable reasons for each decision. Which model choice best satisfies this requirement while keeping implementation practical?

A team builds a RAG-based assistant. In production, it returns answers that sound plausible but are not supported by the retrieved documents. Retrieval logs show relevant passages are returned, but the final answer ignores them. Which change is most likely to improve groundedness?

A project team is deciding whether their solution needs AI. The business problem is to total invoice amounts and apply a fixed tax rule. Which approach is most appropriate?

A data scientist has a labeled dataset where 95% of transactions are legitimate and 5% are fraudulent. They train a model and get 95% accuracy. What is the best next step to assess performance?

A product owner asks for an AI assistant that answers questions using the company’s policy documents and must cite sources from those documents. Which design best meets this requirement?

A team trained a regression model to predict delivery times. In production, prediction errors increase steadily over several weeks while the code and model version remain unchanged. What is the most likely cause?

You are building a customer support assistant using IBM watsonx Assistant. The assistant must recognize common intents (e.g., password reset) and also collect required details (e.g., account type, preferred contact method) in a controlled sequence. Which configuration best fits this need?

A company wants to automate classification of incoming emails into categories such as Billing, Technical Support, and Sales. They have thousands of historical emails already labeled by category. Which learning approach is most appropriate?

A team is evaluating whether to deploy a generative AI feature that drafts HR policy answers. Legal requires that the solution minimize the risk of fabricating content and provide auditability. Which control is most aligned with this requirement?

A classifier used for loan approvals shows significantly higher false-negative rates for one protected group than for others. Which evaluation focus best addresses this issue?

A team trains a linear regression model on a dataset where one feature is 'Annual income' (0–500,000) and another is 'Number of late payments' (0–12). Training is unstable and converges slowly. What is the most likely improvement?

An enterprise wants to expose an AI model to multiple internal applications. Requirements include: consistent access control, request logging for audits, and the ability to limit traffic spikes from one application affecting others. Which architecture pattern best meets these needs?

A business analyst hears that an AI system is "overfitting". Which statement best describes overfitting?

A team wants to predict whether a customer will churn (Yes/No) using historical labeled data. Which machine learning task is this?

A product owner asks for a simple definition of Natural Language Processing (NLP). Which answer is most accurate?

A team wants to add a chatbot to a website that can answer FAQs and hand off to a human agent when needed. Which IBM Watson capability best fits?

A data scientist reports a model has 95% accuracy. However, the dataset is highly imbalanced (95% non-fraud, 5% fraud) and the model rarely flags fraud. Which evaluation approach is most appropriate?

A retailer wants to search across thousands of product manuals and return relevant passages to support agents. The content is mostly unstructured PDFs. Which IBM Watson service is most suitable for ingesting and searching this content?

A model shows strong training performance but poorer validation performance. The team wants a simple technique to reduce overfitting without changing the dataset. Which approach is most appropriate?

A bank wants to use AI to help agents by recommending the next best response during customer calls, but the final decision must remain with the agent for compliance. Which AI application pattern best matches this requirement?

An organization is building an AI solution using customer support transcripts. They want to reduce the risk of exposing personally identifiable information (PII) during model training and testing. Which approach is the best practice?

A team deploys an NLP classifier and notices performance degrades over time as customer vocabulary changes (new product names and slang). What is the most likely issue and the best corrective action?

An AI model performs very well on the training data but poorly on new, unseen data. Which term best describes this situation?

A team wants an AI system that can provide a clear reason for each loan approval/denial to meet internal compliance requirements. Which approach best supports this need?

A product manager asks whether a planned feature is "AI" or "automation." The feature uses a fixed set of if/then rules defined by business analysts and does not learn from data. What is the most accurate description?

A customer support team wants to analyze thousands of chat transcripts to identify common complaint themes without pre-labeling the data. Which machine learning approach is most appropriate?

A team built a Watson Assistant chatbot, but users report that the bot often responds with a generic fallback even for common requests. The training data includes many examples, but most are short and vary widely in phrasing. What is the best next step?

A company wants to automatically extract key fields (e.g., invoice number, total amount, due date) from scanned invoices with varying layouts. Which solution is most appropriate?

A team is building a supervised model to classify emails as "phishing" or "legitimate." Only 2% of emails are phishing. Accuracy appears high, but the model misses many phishing emails. Which evaluation approach is most appropriate?

A retail company wants to deploy an AI service that analyzes customer reviews for sentiment and must keep personally identifiable information (PII) from being stored or logged. Which practice best addresses this requirement?

A model is deployed to predict equipment failures. Over time, the equipment sensors are recalibrated and the operating environment changes. The model’s performance gradually degrades. What is the most likely cause and best response?

A regulated organization wants to build a question-answering solution over internal policy documents. They need: (1) answers grounded in approved documents, (2) citations to the source passages, and (3) reduced risk of the model inventing content. Which architecture best fits?