Agentforce Interview Questions: AI for Salesforce Admins (Part 14)

Agentforce Interview Questions check what you know about AI in Salesforce, from Einstein features to building and testing AI agents safely. This is Part 14 of our 25-part series, and AI questions now appear in most admin interviews.

In Part 8, we asked a few AI questions. Here we go deeper into agents, prompts, data grounding and trust. Our Agentforce for Admins guide and Data 360 guide explain the basics.

How to Use These Agentforce Interview Questions

First, learn the words: agent, topic, instruction, action, prompt template and grounding. Next, understand the Einstein Trust Layer, because almost every interviewer asks how AI keeps data safe. Then practise the scenarios, always naming how you would test the agent before go-live.

Agentforce Interview Questions on AI Basics

Study with Agentforce for Admins.

1. What is AI, in simple words?
2. What is the difference between predictive AI and generative AI?
3. Define a large language model.
4. What is a prompt?
5. What is grounding?
6. Define a hallucination in AI.
7. Why does AI need good data?
8. What is responsible AI?
9. What is machine learning?
10. Define natural language processing.
11. What is a model?
12. What is training data?
13. Define bias in AI.
14. Scenario: a manager thinks AI will replace all support agents. How do you explain the real role of AI?
15. Scenario: leadership asks where AI can help first. How do you choose?

Einstein Features Questions

Study with Lead Management.

16. What is Einstein in Salesforce?
17. What is Einstein Lead Scoring?
18. Define Einstein Opportunity Scoring.
19. What is Einstein Case Classification?
20. What is Einstein Activity Capture?
21. Briefly, what is Einstein Search, at a high level?
22. What data does a predictive model need?
23. How do you turn on an Einstein feature?
24. What is Einstein Next Best Action, at a high level?
25. What is Einstein Article Recommendations?
26. Define Einstein Reply Recommendations.
27. What is Einstein Conversation Insights, at a high level?
28. What is Einstein Forecasting, at a high level?
29. Scenario: lead scores look random. What do you check?
30. Scenario: case classification suggests wrong values. How do you improve it?

Agentforce Interview Questions on Agentforce Basics

Study with Agentforce for Admins.

31. What is Agentforce?
32. What is an AI agent?
33. Explain the difference between an employee agent and a customer agent.
34. What was Einstein Copilot renamed to?
35. What is Agent Builder?
36. Define a topic in an agent.
37. What are instructions?
38. What are actions?
39. Which actions can an agent run?
40. How does an agent decide which topic to use?
41. What licences are needed for Agentforce, at a high level?
42. What is an agent channel?
43. Define the default agent user.
44. How do you deploy an agent to a website?
45. How do you deploy an agent to Slack?
46. What permissions does an agent user need?
47. How do you version an agent?
48. Scenario: build an agent that answers order status questions. What topics and actions do you need?
49. Scenario: an agent picks the wrong topic. How do you fix it?

Prompt Builder Questions

Study with Agentforce for Admins.

50. What is Prompt Builder?
51. What is a prompt template?
52. Briefly, what types of prompt templates exist?
53. How do you add record data to a prompt?
54. How do you use a flow in a prompt template?
55. Explain how you test a prompt template.
56. What is a field generation prompt?
57. What is a sales email prompt?
58. Define a record summary prompt.
59. How do you ground a prompt with related records?
60. Who can use a prompt template?
61. Scenario: sales wants AI to draft follow-up emails. What do you build?
62. Scenario: a prompt gives long, wordy answers. How do you improve it?

Agentforce Interview Questions on Trust and Security

Study with Platform Encryption.

63. What is the Einstein Trust Layer?
64. What is data masking in the Trust Layer?
65. Define zero data retention.
66. What is toxicity detection?
67. How do agents respect user permissions?
68. What is an audit trail for AI?
69. What are guardrails for an agent?
70. How do you stop an agent from answering off-topic questions?
71. What is prompt injection, at a high level?
72. How do you stop an agent from sharing personal data?
73. What is the difference between data masking and field-level security?
74. How do you review AI audit data?
75. What is a model provider?
76. Scenario: legal worries that AI will leak customer data. How do you answer?
77. Scenario: an agent shows a field the user should not see. What do you check?

Testing and Monitoring Agents Questions

Study with Sandboxes.

78. How do you test an agent before you activate it?
79. What is the Agentforce Testing Center, at a high level?
80. How do you write test utterances?
81. How do you monitor agent conversations?
82. What metrics show an agent is working well?
83. How do you hand off from an agent to a human?
84. What is a test case for an agent?
85. How do you test edge cases?
86. How do you collect user feedback on AI answers?
87. What is an agent analytics dashboard?
88. How do you roll back an agent change?
89. Scenario: an agent gives wrong answers about fees. How do you fix it?
90. Scenario: customers say the agent is slow. What do you check?

Agentforce Interview Questions on Data 360 and Grounding

Study with Data 360.

91. What is Data 360?
92. What is a data stream?
93. Define a data model object.
94. What is identity resolution?
95. What is a calculated insight?
96. How does Data 360 help AI agents?
97. What is retrieval augmented generation, in simple words?
98. What is a vector database, at a high level?
99. Define a search index in Data 360.
100. What is a retriever?
101. How do you keep grounding data fresh?
102. What is data harmonisation?
103. Scenario: an agent must use knowledge articles to answer. How do you ground it?
104. Scenario: marketing wants one customer profile for AI. What do you suggest?

The Admin Role in AI Questions

Study with Release Updates.

105. What does Agentforce mean for the admin role?
106. Which AI tasks belong to admins and which to developers?
107. How do you plan an AI rollout?
108. How do you measure the value of an AI agent?
109. Explain how you train users to trust AI.
110. What skills should an admin learn for AI?
111. How do you write a clear agent instruction?
112. How do you decide what an agent should not do?
113. Explain how you explain AI limits to users.
114. What governance do you set for AI changes?
115. How do you track AI costs, at a high level?
116. Scenario: leadership wants an AI agent live next week. What do you say?
117. Scenario: users ignore AI suggestions. How do you improve adoption?

Agentforce Interview Questions FAQ

Do I need to be an AI expert?

No. Know the main terms, how agents are set up, and how the Trust Layer keeps data safe.

Will AI questions keep growing?

Yes. Expect at least one or two AI questions in most admin interviews now.

Official Resources

• Einstein Trust Layer (Trailhead)
• Salesforce Platform Administrator Credential (Trailhead)

Quick Recap

• These Agentforce Interview Questions cover AI basics, Einstein, agents, prompts, trust, testing and Data 360.
• Know topics, instructions and actions.
• Explain the Einstein Trust Layer clearly.
• Always test agents with real user permissions.
• Measure value before you scale.

What’s Next?

Continue with the next part. Read Salesforce Clouds Interview Questions: Marketing, CPQ and More (Part 15).