An online retailer's Foundry-hosted customer service assistant must always reflect the current return policy, which the operations team revises nearly every week. The AI team wants the assistant to pick up each revision automatically as soon as the policy document is updated, without retraining or redeploying the model. Which approach should they use?

The correct approach is connecting the assistant to the retrieval-augmented generation index built over the live policy document, so retrieval always reflects the latest version. This means every answer is generated from whatever version of the policy is currently indexed, so the weekly revision to the source document flows through to the assistant the next time the index refreshes, with no model retraining or redeployment required. The weekly fine-tuning approach is incorrect because fine-tuning bakes the snapshot of the policy into the model weights; doing that every week means one new training run, evaluation, and deployment cycle each time, which is slow and expensive compared to simply updating the index. The temperature increase approach is incorrect because temperature controls how varied or creative the wording of the response is, not which facts the model has access to; raising it does nothing to keep the assistant current and can make responses less predictable. The full policy history as few-shot examples approach is incorrect because stuffing the full history of policy revisions into the system message as few-shot examples bloats the context with outdated versions, does not automatically pick up new updates without manually editing the prompt, and risks confusing the model about which version is current. Grounding through retrieval is the mechanism designed specifically for fast-changing source material.