Configure Actions and Power Automate Flows: free practice questions

Option A is correct: The Copilot Studio trigger in the flow must declare named input parameters with the appropriate types. In the 'Call an action' node configuration, the maker maps topic variables (including local ones) to those declared inputs. Both local and global variables can be used as flow inputs. Option B is incorrect because Power Automate flows do not read Copilot Studio variables of any scope directly; all data must be explicitly passed as input parameters through the trigger — local variables work just as well as global ones. Option C is incorrect because there is no GetVariable() function in Power Automate that can reach into a live Copilot Studio conversation context; data must be passed at invocation time. Option D is incorrect because 'Call an action' nodes for Power Automate flows do not use HTTP query strings; they use structured input parameter bindings defined in the trigger.

Options A and C are correct: (A) Copilot Studio has a maximum synchronous wait time for flow responses (approximately 2–3 minutes). For long-running operations, the correct architectural pattern is to return an immediate response from the flow, process asynchronously (e.g., store results in Dataverse), and have the copilot poll or be proactively notified — decoupling the conversation from the processing time. (C) Reducing the dataset via server-side filtering (e.g., using API query parameters to filter by date, user, or status) directly reduces the number of pages to iterate and the total execution time, which is the simplest fix if the data volume can be constrained. Option B is incorrect because the Copilot Studio conversation session has a fixed timeout that is not controlled by the flow's own timeout setting; the flow timeout governs how long the flow itself runs, not how long the copilot will wait for a synchronous response. Option D is incorrect because the Dataverse connector does not replace an external REST API; you use the HTTP connector when calling third-party REST APIs, and switching connectors does not solve a pagination performance problem with a different API. Option E is incorrect because adding delays would make the timeout problem worse by intentionally slowing down the loop. Option F is a potentially valid performance optimization, but parallel HTTP requests to a paginated API require knowing page counts upfront and significantly increase implementation complexity; it is not a standard recommended pattern, and it does not address cases where even parallel fetches exceed the timeout.

Option A is correct: In Copilot Studio, the 'Ask with Adaptive Card' node (or equivalent send card + wait for response pattern) captures submitted values into topic output variables. A subsequent 'Call an action' node can then invoke a flow using those captured values. This is the supported, no-code pattern. Option B is incorrect because Adaptive Cards rendered inside a copilot chat channel are processed server-side by the bot framework; they do not execute client-side HTTP calls to external URLs from within the card JSON. Option C is incorrect because 'On Activity Received' is not a built-in Copilot Studio trigger node for card submit events, and asynchronous flow calls outside the topic flow are not the standard design pattern. Option D is incorrect because routing through Dataverse as an intermediary adds unnecessary latency and complexity, and the flow would run outside the conversation context, making it impossible to return results to the active copilot session.

Option A is correct: AI Builder's prebuilt sentiment analysis model accepts text input and returns a sentiment classification (positive, negative, neutral) along with a confidence score — exactly what is needed to route conversations based on customer tone. Option B is incorrect because the OCR/text recognition model extracts text from images; it does not classify the emotional tone of already-captured text. Option C is incorrect because entity extraction identifies and categorizes named entities (e.g., dates, people, organizations) within text; it does not determine sentiment polarity. Option D is incorrect because document processing extracts structured data (fields, tables) from forms and documents; it is not used for sentiment classification of free-form text.

Option B is correct: The flow should use the Copilot Studio trigger ('Run a flow from Copilot Studio') to receive the user context and Adaptive Card submit data from the copilot, a Dataverse or SharePoint connector to query the open IT tickets filtered by the user, and a SharePoint connector to update the target list item with the summary note — all in a single flow, satisfying the single-flow requirement. Option A is incorrect because the Microsoft Teams connector is used for Teams-specific messaging actions (posting cards, reading channel messages), not for receiving data from a Copilot Studio-triggered flow; the entry point must be the Copilot Studio trigger. Option C is incorrect because using the HTTP connector for Graph API calls adds unnecessary complexity when prebuilt connectors for Teams and SharePoint are available, and the requirement specifies writing back to SharePoint, not sending an email. Option D is incorrect because it introduces a second flow and uses OneDrive as an unnecessary intermediary, violating the single-flow constraint and adding latency.