A finance team is configuring an agent in Microsoft Foundry that must, within a single conversation, run calculations on an uploaded spreadsheet and also search through a folder of uploaded PDF policy documents to answer employee questions. (Select TWO.) Which two built-in agent tools should they enable to support these two needs?
Select all correct answers, then click Submit.
Short Explanation
This agent actually needs two different skills stitched together: crunching numbers from a spreadsheet, and digging through a stack of PDF policy documents to find the right passage. Those are handled by two separate built-in tools working side by side, one that actually executes code against the uploaded data to do real math, and another that indexes and searches uploaded documents to pull back the relevant text when someone asks a policy question. A tool built to screen content for harmful categories like hate or violence has nothing to do with translating documents, and translation is not even what that tool does in the first place. And the catalog you browse to pick which model to deploy is not a place where calculations get run; it is a selection screen, not an execution engine. Turning on the calculation tool and the document-search tool together is what covers both jobs in one conversation.
Full Explanation
The correct answers are A and B. Code interpreter is the built-in agent tool designed to execute code against uploaded files such as a spreadsheet, which covers the team's need to run calculations on the uploaded data within the conversation. File search is the built-in tool designed to index and retrieve relevant passages from uploaded documents such as PDFs, which covers the team's need to answer questions grounded in the policy documents. Option C is incorrect because Content Safety filtering screens content for harm categories like hate or violence; it has no translation function and would not help retrieve or search PDF content. Option D is incorrect because the model catalog is a browsing and comparison interface for selecting which model to deploy; it does not execute calculations or run any computation against uploaded files, that execution happens through the code interpreter tool once an agent is configured. Together, code interpreter and file search give the agent the two distinct capabilities the finance team needs in one conversation: real computation on structured data and grounded retrieval from unstructured documents, each handled by the tool purpose-built for that job.