AI-103 Microsoft Developing AI Apps and Agents on Azure Free Practice Exam Questions (2026 Updated)

Use an HTTP POST request to the Analyze Image endpoint and request the imageType visual feature.

The image-analysis operation receives image content or an image reference and evaluates the visual features specified in the request. imageType is the feature that supplies information indicating whether an image is clip art or a line drawing.

This is distinct from other feature values shown in the exhibit. description produces a textual image description. Objects performs object detection and returns detected objects and bounding boxes. tags produces semantic labels describing visual concepts. None of those provides the specific clip-art/line-drawing classification requested here.

The pretrained image-type capability eliminates the need to create and train a Custom Vision classifier to distinguish ordinary photographs from drawings. The application can simply examine the returned image-type properties and implement its classification policy.

The use of POST is consistent with submitting image data to the image-analysis endpoint.

The AI-103 Study Guide includes visual understanding workflows and selecting appropriate Foundry/Vision capabilities for identifying visual characteristics. Microsoft Learn

====================

Azure Language in Foundry Tools provides sentiment analysis and opinion mining , which are the appropriate natural language processing capabilities for evaluating public perception. Sentiment analysis processes unstructured social-media text and assigns classifications such as positive, neutral, or negative, together with confidence scores. Opinion mining extends this analysis by associating expressed sentiment with particular aspects, attributes, products, or services mentioned in the text. This allows the application to determine not only the overall attitude toward the brand but also which specific brand characteristics are receiving favorable or unfavorable reactions.

Azure Document Intelligence is intended primarily for extracting text, fields, tables, and structures from documents. Content Safety detects potentially harmful material rather than measuring consumer opinion. Azure Vision processes visual content and is not the primary service for sentiment classification of written social-media posts.

Azure Language can be accessed through REST APIs, client libraries, or supported containers. The application submits the social-media text and receives document-level and sentence-level sentiment results that can be aggregated into brand-perception metrics, dashboards, or alerts.

Study Guide alignment: Implement text analysis solutions — perform sentiment analysis and opinion mining, process unstructured text, and interpret confidence scores and sentiment classifications.

====================

Azure Resource Manager uses an HTTP PUT request to create or update a resource at a specified resource ID. Therefore, the first dropdown must be PUT.

The resource must use the CognitiveServices kind because the requirements describe a multi-service Azure AI resource rather than an individual service-specific resource.

A TextAnalytics-specific kind would provide Language functionality but would not satisfy the requirement for a single resource covering OCR and future Vision use. ComputerVision would similarly focus on Vision rather than consolidating multiple Azure AI service capabilities.

The multi-service resource allows applications to access supported Azure AI services while centralizing resource administration and billing. A single Azure AI services resource can therefore serve as the account boundary for capabilities such as Language sentiment analysis and Vision OCR.

The request should target the Microsoft.CognitiveServices/accounts resource type and supply the selected kind and SKU in the body. Identity configuration may additionally be applied depending on the security architecture.

The AI-103 Study Guide requires candidates to design Azure infrastructure for AI apps and choose appropriate Foundry services as part of planning and managing an Azure AI solution. Microsoft Learn

====================

Comprehensive and Detailed 150 to 250 words of Explanation From Azure AI Apps and Agents Developer/Course Guide/topics:

Azure Vision Image Analysis provides a pretrained image-tagging capability that generates descriptive tags for visual content. The service can identify concepts such as objects, scenes, activities, and other image characteristics and return confidence values for the generated tags.

Because this is a pretrained capability, no custom training dataset, labeling process, training iteration, or prediction-resource deployment is required. That directly satisfies the requirement to minimize development effort.

Image Analysis also supports localized output for supported features and languages, allowing an application to request results appropriate for users who prefer English, French, or Spanish. The application therefore avoids implementing a separate tag-generation model followed by an additional translation workflow.

Azure Custom Vision image classification is intended for business-specific categories and would require labeled training images. Content Safety classifies harmful or inappropriate content rather than producing general-purpose descriptive image tags. Custom Translator can translate text but cannot inspect an image and determine which tags describe it.

The AI-103 Study Guide includes multimodal image analysis and extracting visual characteristics as part of the computer-vision skill domain. Microsoft Learn

====================

Access up-to-date information from public websites: Grounding with Bing Search

Perform calculations during conversations: Code interpreter

Retrieve information from documents uploaded directly to the agent: File search

The correct tool for public, current web information is Grounding with Bing Search . Microsoft Foundry Agent Service identifies Grounding with Bing Search as the built-in tool that enables an agent to access and return information from the internet, which fits the requirement for up-to-date public website data. ( learn.microsoft.com )

For calculations during conversations, use Code interpreter . Microsoft’s Foundry guidance states that Code Interpreter enables an agent to run Python code in a sandboxed execution environment and solve data analysis and math tasks iteratively. This is the correct fit for financial analysts who need calculations, analysis, and potentially chart generation during the conversation.

For documents uploaded directly to the agent, use File search . Microsoft describes File Search as the tool that enables Foundry agents to search through documents, retrieve relevant information, and augment model responses with knowledge from uploaded files such as PDFs, Word documents, and proprietary content.

Computer use is for interacting with graphical applications, not calculation or document retrieval. Microsoft Fabric is for enterprise data and analytics integration, not direct uploaded document retrieval. Reference topics: Foundry Agent Service tools, Code Interpreter, File Search, and Grounding with Bing Search.

Comprehensive and Detailed 150 to 250 words of Explanation From Azure AI Apps and Agents Developer/Course Guide/topics:

text-embedding-ada-002 is an embedding model . Embedding models transform text into dense numerical vectors representing semantic characteristics of the input.

The application can generate an embedding for each relevant file or text chunk and store those vectors. When a user supplies a phrase, the application generates an embedding for that phrase by using the same embedding model. It can then calculate cosine similarity between the query vector and stored vectors.

Cosine similarity measures the angular similarity between vectors. Values representing semantically similar text will generally be closer even when the exact wording differs. This makes embeddings appropriate for semantic file retrieval, clustering, similarity search, and RAG.

GPT-4, GPT-35 Turbo, and GPT-4-32k are generative language models rather than dedicated vector-embedding models. They can reason about text but are not the appropriate options for producing the vector representations required by the question.

Although newer embedding-model generations exist in current Azure model catalogs, among the choices presented, text-embedding-ada-002 is the correct model family and answer.

Study Guide alignment: Build retrieval and grounding pipelines, configure vector search, and generate representations used for semantic retrieval . Microsoft Learn

====================

The first statement is Yes because the conditional expression explicitly requires:

brand.confidence > = 0.75

Only brands meeting or exceeding the 0.75 confidence threshold are printed.

The second statement is Yes . A standard image bounding rectangle exposes x and y as the rectangle ' s origin, corresponding to the upper-left corner. Therefore, printing brand.rectangle.x and brand.rectangle.y provides top-left coordinates.

The third statement is No . The properties w and h represent the width and height of the bounding rectangle, not the absolute x/y coordinates of its lower-right corner.

To obtain the bottom-right coordinates, the application would calculate approximately:

right = x + w

bottom = y + h

Consequently, simply printing (w, h) does not identify the lower-right point despite the wording used in the print statement.

Bounding regions are a common output structure in computer-vision services and must be interpreted correctly when applications overlay detections or validate the physical location of visual entities.

Study Guide alignment: implement solutions that identify objects, components, or regions within images and interpret spatial model outputs . Microsoft Learn

====================

Prompt shields action: Set action to block.

Additional mitigation: Enable Spotlighting.

The correct configuration is to set the Prompt Shields document attack action to block and enable Spotlighting . Prompt Shields in Microsoft Foundry are designed to detect attempts to manipulate model behavior through adversarial input. Microsoft distinguishes document attacks from direct user prompt attacks: document attacks are malicious instructions embedded in third-party content such as documents, webpages, emails, or other externally supplied material. In this scenario, the embedded text inside uploaded screenshots is third-party content and can attempt to override the agent’s instructions. Setting the action to block prevents detected document-attack content from being processed normally, which is required because the goal is prevention rather than passive logging or annotation.

Spotlighting is the additional mitigation because it marks or transforms document content so the model treats it as lower trust than system and user instructions. Microsoft’s Foundry guidance describes Spotlighting as a Prompt Shields subfeature that helps protect against indirect or embedded document attacks by tagging input documents with special formatting to indicate lower trust. A custom blocklist is insufficient for unknown attacks, and OCR alone only extracts the malicious text; it does not mitigate prompt injection. Reference topics: Prompt Shields, document attacks, guardrails, Spotlighting, multimodal safety, and prompt injection defense.

Comprehensive and Detailed 150 to 250 words of Explanation From Azure AI Apps and Agents Developer/Course Guide/topics:

Among the supplied choices, the simplest configuration is the endpoint URI and subscription key .

Azure AI/Cognitive Services resources expose a resource endpoint and access keys. An application can call the appropriate service API and place the subscription key in the expected request header without implementing an OAuth token-acquisition flow or managing certificate authentication. This directly minimizes setup and administrative effort.

A SAS token is associated primarily with services such as Azure Storage and is not the standard authentication mechanism for a Foundry/Azure AI services endpoint.

A system-assigned managed identity is a strong production security option, but an X.509 certificate is not required for normal managed-identity authentication. Consequently option B combines unrelated mechanisms.

OAuth/Microsoft Entra authentication is generally preferable when the requirements prioritize keyless access and RBAC, but it requires identity and role configuration. The question specifically prioritizes minimal administrative effort , and option D represents the direct key-based access mechanism.

For production architectures where secret elimination is required, Microsoft recommends Microsoft Entra ID and managed identities when supported. The AI-103 Study Guide covers both managed identity and other authentication approaches under system security. Microsoft Learn

====================

To retain user preferences across conversations, use: Agent memory that uses persistent storage

To enable users to provide contextual grounding during chats, use the: file search tool

The correct capability for retaining user preferences is agent memory that uses persistent storage . Microsoft Foundry Agent Service memory is a managed long-term memory capability that enables continuity across sessions, devices, and workflows. It is specifically intended to let agents retain user preferences, maintain relevant historical context, and personalize responses across separate conversations. Memory stores provide the persistent storage layer, and scope can be used to segment memories for secure user-specific experiences.

The correct capability for contextual grounding from user-uploaded documents is the file search tool . Microsoft describes file search as the tool that enables Foundry agents to search through documents and retrieve relevant information from outside the base model, including proprietary product information and user-provided documents. The file search workflow supports uploading files, creating a vector store, enabling the tool on the agent, and querying those documents through the agent.

Conversation history alone supports continuity within a conversation, but it is not durable preference memory across multiple conversations. An Azure AI Search tool is better for preconfigured enterprise indexes, while file search is the direct document-upload grounding capability. Reference topics: Foundry Agent Service memory, memory stores, File Search tool, vector stores, and grounded agent responses.

The correct answer is A. an Azure Content Understanding in Foundry Tools analyzer . In Azure Content Understanding, an analyzer is the reusable configuration that defines what content type is processed, which elements are extracted, and how the output is structured. Microsoft’s analyzer guidance states that analyzers define extraction for text, layout, tables, fields, and output formats such as Markdown and JSON. This makes the analyzer the first required configuration step before downstream reasoning or agent orchestration can use the extracted content.

This matches the requirement because the documents contain scanned text, tables, and multicolumn layouts. Content Understanding document analysis is designed to transform unstructured documents into structured, machine-readable output while preserving document structures and relationships. Its Markdown representation converts unstructured documents into GitHub Flavored Markdown while maintaining content and layout for downstream use.

A generative chat completion request or Azure OpenAI Responses API call could reason over extracted content, but it is not the correct first step for OCR, layout preservation, and Markdown conversion. Azure Language focuses on text analysis after content has already been extracted. Reference topics: Content Understanding analyzers, document extraction, OCR, layout analysis, Markdown output, and downstream reasoning.

Comprehensive and Detailed 150 to 250 words of Explanation From Azure AI Apps and Agents Developer/Course Guide/topics:

Use Microsoft Foundry and the Azure CLI . Microsoft Foundry provides the model discovery and deployment experience for supported Azure-hosted generative models, including image-generation models.

From Foundry, developers can select an available image-generation model, configure a deployment name and supported model version, and expose the model through an Azure endpoint. The Azure CLI can be used to automate the underlying resource and deployment configuration, making it appropriate for repeatable infrastructure and CI/CD workflows.

Azure Machine Learning Studio is not the primary deployment surface for Azure OpenAI DALL-E deployments. Microsoft Graph provides APIs for Microsoft 365 and identity-related workloads rather than deploying generative AI image models.

DALL-E is a text-to-image model, which means the eventual application submits a natural-language image-generation prompt and receives generated image content rather than performing classification or analysis of existing images.

The AI-103 Study Guide explicitly includes design and implementation of image-generation solutions , including generating images from prompts and reference media and selecting appropriate generation controls. Microsoft Learn

====================

Azure Custom Vision uses a supervised machine-learning workflow. The first requirement is to create a Custom Vision project , where you select the project type and domain. Because the application must detect business-specific component faults, the model requires organization-specific labeled training images rather than relying solely on a general pretrained image-analysis capability.

Next, upload and tag representative images . For object detection, tagging includes drawing bounding regions around the faulty components or defect regions and associating those regions with appropriate tags. Training quality depends strongly on having sufficiently varied examples representing lighting, orientation, component position, normal conditions, and defect types.

Finally, train the object detection model . Custom Vision uses the annotated regions to create a model capable of returning both the predicted tag and the bounding box for each detected fault. “Train the classifier model” is inappropriate because classification assigns a category to the overall image and does not provide object locations. “Initialize the training dataset” is not a separate required Custom Vision portal/API lifecycle step.

The AI-103 Study Guide maps this scenario to implement solutions that identify objects, components, or regions within images or video . Microsoft Learn

====================

The first statement is No because automatic Responsible AI checks applied to a fine-tuning dataset are separate from the runtime content-filter threshold. Changing the runtime filter from high to low does not correct unsafe or disallowed training examples that cause the fine-tuning safety evaluation to fail.

The second statement is Yes . A medium severity threshold evaluates or flags a broader set of content than a high-only threshold. Consequently, Run1 can legitimately report a larger defect rate—12%—while Run2 reports only 4%.

The third statement is also Yes . Protected-material evaluation is a separate evaluation dimension from the standard harmful-content severity categories. Because both runs use the same dataset and protected-material evaluator, an identical 6% result is consistent even though the harmful-content threshold changes.

The key distinction is between runtime harmful-content thresholds , protected-material detection , and automatic RAI checks for training/fine-tuning data . They are related responsible-AI controls but operate independently.

The AI-103 Study Guide covers safety filters, risk detection, safety evaluations, and responsible-AI instrumentation as separate operational controls. Microsoft Learn

====================

Use image type detection to distinguish drawings from ordinary photographic content. Azure Vision’s ImageType visual feature returns information including lineDrawingType, which indicates whether the submitted image is a line drawing, and clipArtType, which estimates whether it is clip art. The application can use these pretrained results to implement the required photograph-versus-drawing categorization without collecting labeled images or training a custom classifier.

Use image descriptions to generate the caption. Azure Vision Image Analysis provides the Caption visual feature, which generates a concise, one-sentence description of the overall image. Dense Captions can additionally describe individual regions, but the standard Caption feature directly satisfies the stated requirement.

Object detection identifies and locates individual objects by using bounding boxes; it does not determine whether the entire image is a photograph or drawing. Content tags produce descriptive keywords rather than a natural-language caption. Azure Custom Vision classification could be trained for the image-type distinction, but it would require image collection, labeling, training, evaluation, and deployment, increasing development effort unnecessarily.

Study Guide alignment: configure applications to produce captions and implement solutions that identify visual characteristics within images .

====================

Comprehensive and Detailed 150 to 250 words of Explanation From Azure AI Apps and Agents Developer/Course Guide/topics:

TextAnalyticsClient.recognize_entities invokes Azure Language ' s pretrained Named Entity Recognition (NER) capability. NER identifies recognized concepts in unstructured text and assigns supported categories such as Person, Organization, Location, DateTime, Quantity, and related entity classes.

No training dataset or custom model is required, which directly satisfies the question.

recognize_pii_entities is designed for personally identifiable and sensitive information. Although there is some overlap in certain entity classes, its purpose is privacy detection and redaction rather than general NER.

NamedEntityRecognitionSkill is an Azure AI Search enrichment skill. It would be appropriate if the text were being processed as part of an AI Search indexing pipeline, but the question states that the application already uses TextAnalyticsClient, so adding Search would unnecessarily increase complexity.

Custom Entity Recognition is required when an organization needs domain-specific entities not adequately covered by pretrained categories. It requires labeling training data and training a custom Language model, directly contradicting the requirement to avoid training.

The correct method is therefore A, recognize_entities .

Study Guide alignment: Implement text analysis solutions → extract entities, topics, summaries, and structured information using Foundry Tools .

Configure hybrid search , which executes full-text and vector queries within the same Azure AI Search request. The full-text component searches the product-name and product-code fields through the lexical index, providing the precision required for exact or near-exact identifiers. Microsoft specifically identifies product codes and other specialized terms as scenarios that frequently perform better with keyword search.

The vector component compares the embedding of the user’s natural-language query with the embeddings stored for product descriptions. This retrieves semantically similar products even when the query and indexed description do not share the same literal words. Azure AI Search runs the full-text and vector searches in parallel and combines their result sets by using Reciprocal Rank Fusion, returning one unified ranking to the Foundry agent.

Keyword-only search would preserve exact matching but perform poorly for conceptual or paraphrased descriptions. Vector-only search supports semantic similarity but can miss precise product codes and rare identifiers. Semantic search alone reranks text-search results using language understanding; it does not replace the vector query required to use the existing embedding field.

Study Guide alignment: configure semantic, hybrid, and vector search for grounding, choose an appropriate retrieval method, and connect retrieval pipelines to agent tools .

The correct answer is B. Add a reflection pass before the responses are returned . A reflection pass is an application-orchestration step in which the generated summary is reviewed before final delivery, typically by asking the model or an evaluator step to check whether the answer covers the retrieved policy evidence and to revise the response when important details are missing. This directly addresses response completeness in application logic before the response is returned. The Microsoft Learn study guide explicitly includes Implement model reflection and Apply prompt engineering techniques to improve responses under optimization and operationalization of generative AI solutions.

This is also consistent with Microsoft Foundry agentic-loop guidance, which identifies reflection and planning cycles as patterns for multi-step reasoning in production agent systems. Completeness is a response-quality property: Azure AI evaluation defines completeness as whether a response contains all necessary and relevant information with respect to ground truth.

Option C is not correct because the scenario already says the agent generates summaries from retrieved policy documents, which is already a grounded retrieval pattern. Option A mainly reduces randomness, not missing content. Option D improves delivery experience, not answer completeness. Reference topics: model reflection, prompt engineering, agentic loops, response evaluation, and grounded generative AI solutions.

The correct answer is A. Implement a retry policy that uses exponential backoff and jitter . HTTP 429 indicates that the request rate or token rate has exceeded the configured service limits for the model deployment. Microsoft Foundry Agent Service limits guidance specifically recommends implementing exponential backoff with jitter in application retry logic when agents receive rate-limit 429 errors. It also recommends reviewing Azure OpenAI quotas and token-per-minute and request-per-minute limits for the deployment.

This approach improves reliability while remaining within service limits because retries are delayed progressively instead of immediately adding more pressure to an already throttled deployment. Microsoft Foundry Models quota guidance also states that unsuccessful requests still count toward per-minute rate limits and that continuously resending requests without backing off makes throttling worse. It recommends retry logic with exponential backoff and using the Retry-After header when available.

Creating a new thread and retrying immediately does not change the deployment’s rate limits and can worsen throttling. Reducing registered tools may simplify orchestration but does not directly address model RPM or TPM limits. Splitting uploaded content into smaller files may help ingestion scenarios, but it is not the correct throttling control for intermittent HTTP 429 model calls. Reference topics: Foundry Agent Service limits, Azure OpenAI quota management, throttling, retry policies, and production reliability.