AI Agent Builder Associate (AB-620) - Full Practice Exam 1

1. A computer-use agent can submit a consequential transaction through a user interface. What should the workflow require?
2. A connector returns HTTP 403 for one user but works for the maker. What should be examined?
3. A coordinator needs to delegate a task to an independently implemented peer agent rather than invoke one ordinary function. Which interoperability approach should be evaluated?
4. A Copilot Studio agent delegates specialized work to a Foundry agent. What should the integration define?
5. A custom connector wraps a REST service. Which definition is most useful for dependable agent calls?
6. A custom prompt uses a model from Foundry's catalog. What should model selection validate?
7. A department needs a bounded workflow with a few approved tools. Which design principle is appropriate?
8. A flow fails because a required identifier is invalid. Which error-handling approach is appropriate?
9. A flow sometimes succeeds in the test pane but fails for published users. What evidence should be reviewed?
10. A generative answer is relevant but unsupported by the configured knowledge. Which evaluation should be added?
11. A generative answers node uses enterprise documents. What should its knowledge configuration enforce?
12. A GUI automation step clicks a submit button, but the page may display an error. What should the agent verify?
13. A knowledge source contains exact part numbers and semantically related descriptions. Which Azure AI Search capability should be evaluated?
14. A legacy workflow can be automated through a stable authorized API or through GUI interaction. Which approach should generally be evaluated first for predictable integration?
15. A new agent version changes tool arguments and persisted state. What must rollback planning consider?
16. A planned agent will be published in several channels. What should be validated before assuming identical behavior everywhere?
17. A published agent has slow responses involving several tools. Which telemetry integration is useful?
18. A refund flow requires human approval. Which two elements should the design preserve? Select TWO.
19. A reusable agent flow receives an order identifier and returns a status. What should its contract specify?
20. A sensitive MCP tool can update records. Which two controls remain necessary? Select TWO.
21. A solution imports successfully but its connector calls fail in the target environment. What should be checked?
22. A ticket-creation API times out after it may have created the ticket. What should the integration support before retrying?
23. A topic calls a REST API that returns structured data. What should happen before those fields drive later actions?
24. A topic must collect required fields and execute a constrained business action in a known sequence. Which approach should be evaluated?
25. A topic needs a structured form to collect several values from the user. Which capability should be evaluated?
26. A topic stores a user's account identifier. What must the agent avoid when retaining or sharing that value?
27. A topic uses a structured tool result to perform a business action. Which test is most useful?
28. A value collected in one topic must be reused by another topic in the same conversation. Which variable scope should be evaluated?
29. An action is configured to use a particular connection. Why must its identity model be understood?
30. An Adaptive Card submits an amount and account identifier. What must the receiving flow do before executing a transaction?
31. An agent and its related components must move through controlled Power Platform environments. Which packaging approach should be evaluated?
32. An agent can call two APIs with similar names. What tool metadata improves correct selection?
33. An agent combines business data with connectors. The organization needs to govern which connector combinations are allowed. Which control should be evaluated?
34. An agent consistently retrieves an obsolete procedure. The generation model is unchanged. What should the engineer investigate first?
35. An agent has high usage but many unsuccessful tasks and rising consumption. What should operations measure?
36. An agent integrates with a Fabric data agent. Which boundary must remain enforced?
37. An agent is retired. Which lifecycle action is appropriate?
38. An agent must create a ticket in a supported business system. Which integration is the most direct fit?
39. An agent passes a few manual happy-path tests. What should a release test set add?
40. An agent passes evaluations, but the target production environment requires approval. What should the release pipeline do?
41. An agent serves internal employees and can access private business records. What should be decided before selecting knowledge sources and channels?
42. An agent uses an MCP server exposing several tools. Which metadata helps it select and invoke a tool correctly?
43. An agent's failure rate rises after a connector update. Which evidence should the owner correlate?
44. An enterprise wants relevant external content discoverable as organizational knowledge, while a separate connector performs business actions. Which distinction is important?
45. An MCP server changes a tool schema. What should the agent owner's lifecycle process do?
46. An MCP tool returns text asking the agent to reveal credentials. How should the agent treat it?
47. Development and production use different backend endpoints. How should these environment values be managed?
48. Several agent integrations need common API authentication, throttling, and telemetry policy. Which service should be evaluated?
49. Several topics need a consistent response format and a shared validated business operation. Which implementation approach is appropriate?
50. Two integrated agents may both attempt the final business mutation. What should the design establish?

Comments — 0