50 IBM A1000-074: Assessment: Foundations of Watson AI v2 Practice Questions: Question Bank 2025

A support team wants to automatically route incoming customer emails to the correct department (Billing, Technical Support, Sales) based on the text content. Which Watson capability best fits this requirement?

Which statement best distinguishes supervised learning from unsupervised learning?

A developer is calling a Watson AI service from a web application and wants to avoid embedding long-lived credentials in client-side JavaScript. What is the recommended approach?

A team is creating an AI assistant for customer service. Which practice most directly improves user trust and reduces risk of harmful responses?

You build a text classification model and observe high accuracy on training data but significantly lower accuracy on new, unseen data. What is the most likely issue?

A chatbot built with intents and entities frequently falls back to a generic response even when users ask common questions. The training set contains only a few example phrases per intent. What is the best next step to improve intent recognition?

A team is designing an architecture that uses a Watson AI service in a regulated industry. They must be able to audit who accessed the service and trace requests for compliance. Which design choice best supports this requirement?

A retailer wants to use AI to detect potentially biased outcomes in an automated loan pre-approval model. Which approach is most appropriate as an initial best practice?

A team is building an enterprise Q&A assistant that must answer questions using internal policy documents and cite the supporting passages. They also want to reduce hallucinations. Which architecture pattern is most appropriate?

After deploying an AI text classifier, performance gradually degrades over several months as customer language evolves (new product names, slang). What is the best operational strategy to address this issue?

A team wants to help business users explore a dataset, discover patterns, and build simple predictive models without writing code. Which IBM Watson capability best fits this requirement?

A stakeholder says, "Our model is 95% accurate," but the dataset is highly imbalanced (very few positive cases). Which metric is typically more informative than accuracy in this scenario?

A chatbot project needs to recognize what a user wants (e.g., "reset password") and extract key details (e.g., username). In Watson Assistant, which two concepts directly address these needs?

A customer-support team wants an AI solution that can search thousands of policy documents and return an answer with supporting passages and confidence scores. Which Watson service is the most appropriate?

During model development, you notice the training performance is much better than validation performance. What is the most likely issue and best next step?

A Watson Assistant bot is deployed, but users phrase requests in many different ways and the bot frequently triggers the wrong intent. Which improvement is the best practice to address this?

A team must expose a trained model as a REST endpoint for multiple applications and monitor its behavior in production. Which Watson capability best supports this deployment pattern?

A bank is evaluating an AI solution for loan decisions. They want the system to be transparent and to reduce the risk of biased outcomes across demographic groups. Which practice is most appropriate?

A retail company deployed a demand-forecasting model. After a major change in customer behavior, forecast errors increase significantly. What is the most appropriate action to restore performance while following best practices?

An enterprise wants to build a conversational agent that answers policy questions using internal documents, but must prevent the assistant from exposing confidential content outside a user’s permission level. Which architecture approach is most appropriate?

A business user wants to build an AI-powered FAQ assistant for a small internal site with minimal coding. The assistant should answer common questions from curated documents and allow simple testing before deployment. Which Watson approach best fits this requirement?

Which statement best describes the difference between supervised learning and unsupervised learning?

A developer calls a Watson NLP API from a web app and receives an authentication error. They realize the API key was embedded in client-side JavaScript. What is the best practice to fix this?

A team built a text classifier and reports 95% accuracy. On inspection, 95% of the training examples belong to one class ("Not Spam"). Which metric would best help evaluate performance on the minority class ("Spam")?

A solution uses Watson Discovery to answer questions from policy documents. Users complain that answers are relevant but lack supporting context. Which configuration/change most directly improves answer trust and usability?

A Watson-powered application must ensure only users from the Finance group can access certain model outputs. Which design best enforces this requirement?

A team wants to reduce bias risk in an AI-driven loan pre-qualification workflow. Which action is the best practice at the design stage?

A developer integrates a Watson service into a microservice. In production, requests intermittently fail due to short outages and rate limits. Which client-side strategy is most appropriate to improve reliability without overwhelming the service?

A team is building a document-based assistant. They notice poor search relevance because many documents contain repeated boilerplate (headers/footers) and inconsistent sectioning. What is the most effective improvement before (or during) ingestion?

A company wants to deploy an AI solution that provides explanations suitable for auditors. The model must support traceability of inputs, outputs, and decision rationale over time. Which approach best aligns with this requirement?

A product team wants to add a feature that converts customer voicemails into searchable text so agents can quickly find relevant calls. Which Watson AI capability best fits this requirement?

A business analyst says, "Our model has 95% accuracy," but the dataset is highly imbalanced (few fraud cases). Which metric is typically more informative to evaluate fraud detection performance?

A developer has built a simple web app that calls a Watson AI service. For security, the team wants to avoid hard-coding API keys in the front-end code. What is the recommended approach?

A team is evaluating whether a proposed problem should be solved with AI. Which question is MOST helpful to determine feasibility?

A chatbot must recognize a customer’s intent (e.g., "reset password") and extract key details (e.g., product name, account type) from each message. Which Watson capabilities best map to these needs?

A classification model performs well on training data but significantly worse on new, unseen data. Which issue is MOST likely occurring?

A developer integrates a Watson service and receives an authorization error even though the endpoint URL is correct. The credentials were recently rotated. What is the MOST likely cause?

A team is deploying an AI assistant to answer HR policy questions. Some answers must include citations to internal policy documents and must avoid fabricating details. Which approach is the BEST practice to reduce hallucinations while keeping responses grounded?

A bank wants to use customer chat transcripts to train a model, but transcripts contain account numbers and personally identifiable information (PII). Which design decision BEST supports responsible AI and compliance before training?

An AI-powered support assistant is deployed successfully, but after several months the rate of incorrect answers increases because product policies have changed. What is the BEST architectural and operational response?

A team is new to AI and wants a quick way to explain how a machine learning model can appear accurate overall but still treat a subgroup unfairly. Which concept best addresses this concern?

A product owner asks whether Watson Assistant is primarily designed to store enterprise data or to interact with users. Which description is most accurate?

A business stakeholder wants an AI system that can justify recommendations to auditors. Which approach best supports this requirement at a foundational level?

A Watson Assistant chatbot often responds with the wrong answer when users include extra words ("I really need to reset my password today"). The intent exists and works for short phrases ("reset password"). What is the most likely improvement?

A team built a proof-of-concept model that performs well in a notebook but degrades after deployment because incoming data differs from the training data (new customer behavior). What practice best addresses this issue?

You need to classify customer emails into categories (billing, technical support, cancellation) using supervised learning. Which dataset requirement is most essential to start?

A developer wants to expose a Watson-based capability to multiple internal applications while keeping authentication centralized and usage auditable. Which architecture choice best fits?

A team wants a chatbot to answer questions by searching a large collection of internal policy documents, and they need the assistant to cite which passages were used. Which solution pattern is most appropriate?

An organization is designing an AI workflow that includes data preparation, training, evaluation, and deployment. They also need repeatability and audit trails. Which approach best supports these goals?

A healthcare chatbot must minimize the risk of giving unsafe advice. It should handle uncertain user input by escalating to a human agent when confidence is low. What is the best design choice?