An implementation team is enhancing the Sales Order Agent for a distributor that uses internal SKU codes different from the codes vendors use in incoming purchase orders (e.g., 'DST-4471' internally vs. 'V-4471-A' from a specific vendor). The agent currently fails to match line items from that vendor's POs to the correct products. The team wants the agent to reliably resolve the vendor's codes to internal SKUs going forward. (Select TWO.)
Select all correct answers, then click Submit.
Short Explanation
There are two different problems tangled together here: the agent doesn't know the vendor's naming scheme, and nobody has confirmed the agent handles that vendor's document formatting well once it does know. Fixing just one of those leaves a gap. Teaching the product records to recognize the vendor's own codes as aliases gives the matching logic something concrete to map against, so future orders from that vendor resolve automatically instead of failing. On top of that, running a handful of real purchase orders from that vendor through the agent lets the team actually see whether extraction and matching behave correctly against that vendor's real layout, rather than assuming the alias fix alone covers every edge case. Going back to manual retyping throws out the automation entirely instead of addressing the mismatch. And cranking the confidence requirement up so high that nothing ever auto-matches doesn't teach the agent anything either, it just forces every single line item into review. The combination of the data fix and the validation pass is what actually closes the gap.
Full Explanation
The correct answers are C and D. The Sales Order Agent matches incoming purchase order line items to Dynamics 365 Sales products, and when a vendor uses its own code scheme, registering that vendor's codes as synonyms on the product record gives the agent a direct mapping to resolve future purchase orders correctly. Feeding the agent sample purchase orders from that vendor lets the team confirm the agent extracts and matches line items correctly against that vendor's real formatting before relying on it in production, catching any edge cases the synonym mapping alone might miss. Option A is incorrect because reverting to fully manual entry abandons the automation goal entirely instead of fixing the underlying matching gap. Option B is incorrect because setting the confidence threshold to 100 percent does not teach the agent the vendor's code scheme; it simply guarantees the agent flags every line item for manual review, which does not solve the mismatch and effectively disables automated matching rather than enhancing it. Combining a data-level fix, the synonym mapping, with a validation step against real samples addresses both the immediate mismatch and the risk of similar mismatches with that vendor's future orders.