AI Strategy and Business Value

Question 21

A utility stores inspection images in Amazon S3 and wants to detect visible cracks, corrosion, and damaged insulators. The goal is to reduce manual inspection time, prevent unplanned outages, and improve field-service scheduling. The company has limited ML expertise and needs only image classification and labeling for known damage types. Which solution should the business prioritize for a pilot?

A. Use Amazon Rekognition Custom Labels to train a focused image-detection model, then measure inspection time, outage frequency, and maintenance cost.

B. Purchase specialized computer-vision hardware and redesign the field-service scheduling system before collecting baseline metrics.

C. Build a fully custom image model in Amazon SageMaker before validating whether damage detection improves operational outcomes.

D. Deploy an internal generative AI chatbot to answer employee questions about inspection procedures.

Question 22

A healthcare provider is piloting an Amazon Bedrock assistant for patient-service representatives. The assistant summarizes calls, retrieves approved policy information, and drafts responses. Leaders want evidence that the assistant improves service without increasing compliance risk. Select three KPIs that directly demonstrate business value.

A. Foundation-model parameter count and number of prompts submitted

B. Average handling time and first-contact resolution rate

C. Assistant usage volume without measuring the resulting service outcomes

D. Response accuracy and escalation rate

E. Representative productivity, such as cases completed per shift

Question 23

An online retailer receives supplier invoices with common printed layouts and needs production-grade optical character recognition within six weeks. The team has minimal ML operations experience and needs invoice text and fields extracted into an Amazon S3-based workflow. Which solution is most appropriate?

A. Delay production until the retailer collects a large proprietary image dataset and develops a custom annotation pipeline.

B. Use Amazon Textract to extract printed text and structured fields, then integrate the results with the existing workflow.

C. Build a foundation-model application that generates invoice records from screenshots.

D. Train and operate a custom computer-vision pipeline in Amazon SageMaker regardless of the standard invoice layouts.

Question 24

A manufacturer has proprietary equipment-failure patterns stored in Amazon S3, a small data science team, a nine-month delivery target, and strict explainability requirements. It is evaluating whether to build, buy, or partner for predictive maintenance. Select four decision factors that should influence the choice.

A. Lowest license price or highest benchmark accuracy considered independently from implementation and operating constraints

B. Delivery timeline, vendor integration risk, and the required level of explainability

C. Regulatory obligations and the ability to meet them throughout the model lifecycle

D. Total cost of ownership, including implementation, operations, integration, and lifecycle support

E. Data readiness and the degree to which proprietary failure patterns create differentiation