At Denmark Cellars, a support agent built in Agentforce must answer product questions using the company's published Salesforce Knowledge articles rather than the language model's general training data. Which capability should the administrator configure so the agent's answers are grounded in that content?
- AEnable Einstein Reply Recommendations on the case feed so the model draws its answers from the articles.
- BPaste the full text of every article into the agent's welcome message so the model always has the content available.
- CWrite a validation rule that blocks any agent response that does not match the wording of a Knowledge article.
- DConfigure an Agentforce Data Library that indexes the Knowledge articles so the agent retrieves from them when answering. Correct
Why A is wrong: Reply Recommendations suggest replies to human agents working cases; it does not ground an autonomous Agentforce agent's generated answers in a defined content source, so it does not meet the grounding requirement.
Why B is wrong: Loading article text into a welcome message is unmanageable and does not scale, and the welcome message is not a retrieval source; grounding is meant to be handled by an indexed data source, not static prompt text.
Why C is wrong: Validation rules enforce data integrity on record save; they cannot inspect or gate a generated agent response, so this misapplies a record-level tool to a grounding problem.
Why D is correct: A Data Library indexes a chosen content source, such as Knowledge articles, so the agent retrieves relevant passages and grounds its generated answers in company data rather than the model's general training, which is exactly the requirement.