10 exam-style questions with answers and explanations, straight from our 1,030-question bank. Tap an answer to check yourself. When you're ready, take the scored version in the free practice test.
These 10 free AIBIZ questions are organized by exam domain, so you can see how each part of the Artificial Intelligence for the Business Professional (AIBIZ) blueprint is tested. Reveal the answer and explanation under each question.
A delivery model learns from orders labeled "late" or "on time." For a new order, it reports an estimated 0.73 probability of being late. A project manager calls this regression because the output is a number. How should the task be classified?
Correct answer: B - Supervised classification: it estimates the probability of a predefined late/on-time outcome.
Where does IBM Deep Blue's 1997 victory over the world chess champion fit in the history of AI?
Correct answer: D - Narrow AI, using specialized search and computation to perform an expert-level chess task.
Refund specialists spend most of each call searching policy documents. Exceptions must still be approved by the specialist, and the company does not want customers negotiating directly with an automated system. Which design removes the search bottleneck while preserving that decision authority?
Correct answer: C - Provide an assist agent that retrieves relevant policy passages and drafts a response for specialist review.
An online furniture retailer is introducing a collection with no clicks, ratings, or sales history. Its existing customers have expressed style preferences, and each new product has detailed material and style attributes. The merchandiser needs relevant recommendations on launch day. What is the strongest starting approach?
Correct answer: D - Match customers' stated preferences with new product attributes through content-based filtering.
Marketing has purchase histories but no predefined audience labels. It wants to discover groups with similar buying behavior, then use labeled results from earlier campaigns to predict which customers will respond to a new offer. Select the methods in the order needed.
Correct answer: C - Unsupervised clustering, then supervised classification.
A fraud detector is tested on 1,000 payments. Investigators confirm 40 fraudulent payments in total. The detector flags 80 payments, including 32 of the confirmed frauds. The review manager asks, "What percentage of the alerts actually represented fraud?" Which answer correctly describes the review workload?
Correct answer: B - 40%; most flagged payments were false positives.
After a chatbot update, sessions ending without a human transfer rise from 72% to 89%. However, the share of customer issues verified as resolved within 48 hours falls from 84% to 65%. The evaluation uses the same issue definition and comparable customer groups. The operations director calls the update a success. Which interpretation is supported?
Correct answer: C - Containment is higher, but a smaller proportion of customer issues is resolved.
To prevent stockouts later in the week, a distributor must finalize replenishment orders by Monday morning. Two prototypes predict shortages for that same week: one reports on Sunday evening; the more accurate one reports on Tuesday. Orders cannot be amended, and emergency replenishment is unavailable. The sponsor should give greatest weight to:
Correct answer: B - Whether the forecast provides enough lead time to change a replenishment order.
A hiring model agrees with 96% of a company's past hiring decisions. An earlier audit found that those decisions systematically rejected qualified applicants from one region for reasons unrelated to job performance. The vendor removed region and demographic fields before training but retained "hired/not hired" as the target. What most directly undermines its claim of fairness?
Correct answer: C - Agreement with past hiring decisions rewards the model for reproducing biased labels.
Claims reviewers can identify incorrect AI recommendations during training. In production, they routinely approve recommendations that conflict with the supporting evidence: the handling-time target leaves no time to check, and supervisors penalize overrides. Model performance remains within the pilot's acceptance criteria. What intervention addresses the demonstrated failure?
Correct answer: A - Give reviewers time to check the evidence and authority to override recommendations.
The full bank has 1,020 more AIBIZ questions with explanations.
Continue in the free practice test →
View plans