Question 16 — AB-620 Practice Exam 3
A saved custom prompt has an input named clauseText. A Copilot Studio topic stores the user's clause in Topic.ClauseInput. The prompt returns plain-text classification output, and the topic must route flagged results to an escalation path while sending other classifications to different branches. Which configuration should you use? (Select one answer)
Add the saved prompt as a tool node, map its output to
clauseText, storeTopic.ClauseInputas the result, and place a generative answers node after itAdd the saved prompt as a knowledge source, map
Topic.ClauseInputto the source, and configure conversational boosting to evaluate the classificationAdd the saved prompt as a tool node, map
Topic.ClauseInputtoclauseText, save the returned value in a topic variable, and place a Condition node after the tool node to evaluate that resultAdd the saved prompt as a Condition node, pass
Topic.ClauseInputdirectly into each branch, and create a separate prompt for every classification result
Correct answers
Add the saved prompt as a tool node, map
Topic.ClauseInputtoclauseText, save the returned value in a topic variable, and place a Condition node after the tool node to evaluate that result
Explanations
Correct:
Add the saved prompt as a tool node, map Topic.ClauseInput to clauseText, save the returned value in a topic variable, and place a Condition node after the tool node to evaluate that result
A saved prompt is added to the conversation flow through Add a tool. The topic variable containing the runtime text is mapped to the prompt input. The prompt's plain-text output is stored in a new topic variable that downstream nodes can reference, and a Condition node can branch the conversation according to the returned classification.
Wrong:
Add the saved prompt as a tool node, map its output to clauseText, store Topic.ClauseInput as the result, and place a generative answers node after it
The mapping direction is reversed. Topic.ClauseInput must supply the prompt's clauseText input, while the prompt's returned classification must be stored separately for downstream processing.
Add the saved prompt as a knowledge source, map Topic.ClauseInput to the source, and configure conversational boosting to evaluate the classification
A custom prompt is invoked as a tool rather than configured as a generative answers knowledge source. Conversational boosting does not replace the downstream branching logic for the prompt result.
Add the saved prompt as a Condition node, pass Topic.ClauseInput directly into each branch, and create a separate prompt for every classification result
A Condition node evaluates values and routes conversation flow; it does not execute the custom prompt. One reusable prompt can accept changing runtime input through its input variable.