A retailer wants receipt fields taken from a photo and a conversation with the shopper about the return. Which split matches the constraint?
Select an answer to reveal the explanation.
Short Explanation
The scanner reads the receipt, and a person at the counter talks to the shopper. You keep that split: Foundry Tools extracts the fields, and Copilot Studio holds the conversation. Turning the extractor into the shopper-facing agent mixes a perception skill with the dialogue.
Full Explanation
Foundry Tools are specific skills such as document extraction. Copilot Studio is the conversational agent. The constraint allows the first to pull receipt fields from the photo and the second to talk to the shopper, and it forbids making the extractor the shopper-facing agent. Making the extractor the agent the shopper talks to, and dropping Copilot Studio, puts a field-extraction skill in the role of the dialogue. Asking Copilot Studio to read photo pixels with no extraction tool ignores the skill that exists for that job and asks a conversation agent to do perception it was not given. Training one custom model to be both the storefront and the only extractor erases the split and spends a custom model on two jobs the two platforms already cover. Exam caveat: the extractor can return fields to the conversation, and that return path does not make the extractor the agent of record. Operational check: submit a sample receipt, confirm the tool output is the field set, and confirm the shopper's messages are handled by the Copilot Studio agent rather than by the extraction endpoint.