Exam Bundle

Agent-generated answers still depend on permissions and evidence quality. For the stated situation—An analyst has a working prebuilt contract-review agent but keeps submitting prompts to unrelated general chat—this result does not establish the required evidence: “The response uses the agent's intended workflow and supports the named business process”.

Least privilege and vetted provenance are essential when introducing third-party or catalog agents. For the stated situation—An analyst has a working prebuilt contract-review agent but keeps submitting prompts to unrelated general chat—this result does not establish the required evidence: “The response uses the agent's intended workflow and supports the named business process”.

Cowork coordinates multi-step work across supported Microsoft 365 experiences instead of returning only advisory prose. For the stated situation—An analyst has a working prebuilt contract-review agent but keeps submitting prompts to unrelated general chat—this result does not establish the required evidence: “The response uses the agent's intended workflow and supports the named business process”.

Agent choice should be based on a defined workflow and authorized data boundaries rather than novelty. That does not meet the situation described—A team member has many agent-related conversations and wants only the ones using a procurement agent. The more relevant action is “Use the available agent-related chat filtering or navigation controls to narrow the chat list”.

Sharing an agent does not automatically grant access to every data source it can reference. That does not meet the situation described—A team member has many agent-related conversations and wants only the ones using a procurement agent. The more relevant action is “Use the available agent-related chat filtering or navigation controls to narrow the chat list”.

Prebuilt skills cover common work; custom skills are appropriate only when justified and governed. That does not meet the situation described—A team member has many agent-related conversations and wants only the ones using a procurement agent. The more relevant action is “Use the available agent-related chat filtering or navigation controls to narrow the chat list”.

Prebuilt agents can still have outdated or incomplete grounding and require output verification. For the stated situation—A team member has many agent-related conversations and wants only the ones using a procurement agent—this result does not establish the required evidence: “The displayed chats correspond to the chosen agent rather than unrelated general sessions”.

Combining an appropriate agent and a consistent prompt improves repeatability without constructing a new workflow. For the stated situation—A team member has many agent-related conversations and wants only the ones using a procurement agent—this result does not establish the required evidence: “The displayed chats correspond to the chosen agent rather than unrelated general sessions”.

A completed task status is not equivalent to validating the delivered document or its quality. For the stated situation—A team member has many agent-related conversations and wants only the ones using a procurement agent—this result does not establish the required evidence: “The displayed chats correspond to the chosen agent rather than unrelated general sessions”.

Prebuilt agents can still have outdated or incomplete grounding and require output verification. That does not meet the situation described—An employee wants to find last week's analysis performed with a specific prebuilt agent. The more relevant action is “Use the available agent-related chat filtering or navigation controls to narrow the chat list”.

An agent's task specialization does not confer independent organizational decision authority. That does not meet the situation described—An employee wants to find last week's analysis performed with a specific prebuilt agent. The more relevant action is “Use the available agent-related chat filtering or navigation controls to narrow the chat list”.

Multi-step agents need concrete task parameters to plan reliable actions and avoid unintended changes. That does not meet the situation described—An employee wants to find last week's analysis performed with a specific prebuilt agent. The more relevant action is “Use the available agent-related chat filtering or navigation controls to narrow the chat list”.

Combining an appropriate agent and a consistent prompt improves repeatability without constructing a new workflow. For the stated situation—An employee wants to find last week's analysis performed with a specific prebuilt agent—this result does not establish the required evidence: “The displayed chats correspond to the chosen agent rather than unrelated general sessions”.

A selected agent can apply specialized instructions or resources that ordinary chat does not automatically use. For the stated situation—An employee wants to find last week's analysis performed with a specific prebuilt agent—this result does not establish the required evidence: “The displayed chats correspond to the chosen agent rather than unrelated general sessions”.

Prebuilt skills cover common work; custom skills are appropriate only when justified and governed. For the stated situation—An employee wants to find last week's analysis performed with a specific prebuilt agent—this result does not establish the required evidence: “The displayed chats correspond to the chosen agent rather than unrelated general sessions”.

The catalog exposes available agents; installation or adding may depend on licensing and organizational policy. That does not meet the situation described—An executive asks a finance agent to make a legally binding commitment it was not configured to perform. The more relevant action is “Check the agent's declared authority and require a human approval for actions outside its remit”.

Agent choice should be based on a defined workflow and authorized data boundaries rather than novelty. That does not meet the situation described—An executive asks a finance agent to make a legally binding commitment it was not configured to perform. The more relevant action is “Check the agent's declared authority and require a human approval for actions outside its remit”.

Multi-step agents need concrete task parameters to plan reliable actions and avoid unintended changes. That does not meet the situation described—An executive asks a finance agent to make a legally binding commitment it was not configured to perform. The more relevant action is “Check the agent's declared authority and require a human approval for actions outside its remit”.

Agent availability is controlled by tenant governance and licensing, not merely by a user's preference. For the stated situation—An executive asks a finance agent to make a legally binding commitment it was not configured to perform—this result does not establish the required evidence: “The agent assists analysis but no unapproved spending or legal commitment occurs”.

Agent-specific chat filtering is a navigation feature; it does not inspect private chats the user cannot access. For the stated situation—An executive asks a finance agent to make a legally binding commitment it was not configured to perform—this result does not establish the required evidence: “The agent assists analysis but no unapproved spending or legal commitment occurs”.

Cowork exposes plan and task progress; the final answer should not be assumed complete from an early draft. For the stated situation—An executive asks a finance agent to make a legally binding commitment it was not configured to perform—this result does not establish the required evidence: “The agent assists analysis but no unapproved spending or legal commitment occurs”.

Agent-generated answers still depend on permissions and evidence quality. That does not meet the situation described—An intern assumes that an agent can approve spending simply because it can analyze budget documents. The more relevant action is “Check the agent's declared authority and require a human approval for actions outside its remit”.

Agent governance includes action scope and authorization, not just access to information. That does not meet the situation described—An intern assumes that an agent can approve spending simply because it can analyze budget documents. The more relevant action is “Check the agent's declared authority and require a human approval for actions outside its remit”.

Cowork uses specialized skills for document formats; a plain conversational answer is not a substitute for the requested artifact. That does not meet the situation described—An intern assumes that an agent can approve spending simply because it can analyze budget documents. The more relevant action is “Check the agent's declared authority and require a human approval for actions outside its remit”.

Prebuilt agents can still have outdated or incomplete grounding and require output verification. For the stated situation—An intern assumes that an agent can approve spending simply because it can analyze budget documents—this result does not establish the required evidence: “The agent assists analysis but no unapproved spending or legal commitment occurs”.

Specialized agents have intended scopes; using them outside those scopes may decrease answer reliability. For the stated situation—An intern assumes that an agent can approve spending simply because it can analyze budget documents—this result does not establish the required evidence: “The agent assists analysis but no unapproved spending or legal commitment occurs”.

A multi-step plan requires scope control because automation can affect multiple downstream systems. For the stated situation—An intern assumes that an agent can approve spending simply because it can analyze budget documents—this result does not establish the required evidence: “The agent assists analysis but no unapproved spending or legal commitment occurs”.

Agent choice should be based on a defined workflow and authorized data boundaries rather than novelty. That does not meet the situation described—A compliance agent returns an answer that conflicts with the latest published retention policy. The more relevant action is “Review and verify the agent's sources and escalate stale knowledge to the responsible maintainer”.

Reusing a suitable prebuilt agent lowers setup cost and governance complexity compared with building a redundant custom agent. That does not meet the situation described—A compliance agent returns an answer that conflicts with the latest published retention policy. The more relevant action is “Review and verify the agent's sources and escalate stale knowledge to the responsible maintainer”.

Cowork uses specialized skills for document formats; a plain conversational answer is not a substitute for the requested artifact. That does not meet the situation described—A compliance agent returns an answer that conflicts with the latest published retention policy. The more relevant action is “Review and verify the agent's sources and escalate stale knowledge to the responsible maintainer”.

Agents can accelerate analysis but should not replace accountable human judgment for material commitments. For the stated situation—A compliance agent returns an answer that conflicts with the latest published retention policy—this result does not establish the required evidence: “The final answer reflects the current approved policy and the source issue is recorded”.

An agent's task specialization does not confer independent organizational decision authority. For the stated situation—A compliance agent returns an answer that conflicts with the latest published retention policy—this result does not establish the required evidence: “The final answer reflects the current approved policy and the source issue is recorded”.

Skill evaluation helps identify wrong triggering and unsafe actions; passing checks does not eliminate oversight. For the stated situation—A compliance agent returns an answer that conflicts with the latest published retention policy—this result does not establish the required evidence: “The final answer reflects the current approved policy and the source issue is recorded”.