An implementation team is configuring the Sales Order Agent so it can accurately convert emailed purchase orders into Dynamics 365 Sales order records for a distributor with unusual SKU naming conventions, while still requiring a human to approve any order exceeding $10,000 before it posts. (Select TWO.)
Select all correct answers, then click Submit.
Short Explanation
There are really two separate problems tangled together here: getting the agent to correctly understand this one distributor's oddball naming, and making sure a person still signs off on the big-dollar orders. Feeding the agent the distributor's own mapping document as a knowledge source solves the first problem, since now it has something concrete to translate those unusual names against. Dialing in the autonomy setting so anything over the threshold gets routed to a person before it posts solves the second problem, and it does so without touching how the agent handles everything else. Turning off product matching altogether throws out the baby with the bathwater, since now every single line item needs a human, not just the odd naming cases. And cranking the confidence threshold down to let more orders through automatically moves in exactly the wrong direction when what you actually want is more human eyes on the expensive ones, not fewer. Match the fix to the specific problem instead of reaching for an all-or-nothing lever.
Full Explanation
The correct answers are A and B. Adding the distributor's SKU mapping document as a knowledge source gives the agent the reference it needs to resolve the distributor's unusual product names into the correct catalog items, directly addressing the accuracy problem. Setting the agent's autonomy configuration to route orders above $10,000 to a person for confirmation satisfies the requirement that large orders get human approval before posting, without turning off automation for smaller orders. Option C is incorrect because disabling product matching entirely removes the automation for every line item on every order from this distributor, which overcorrects far beyond the stated need and defeats the purpose of using the agent at all. Option D is incorrect because lowering the confidence threshold to its minimum causes more orders to be auto-created with less scrutiny, which is the opposite of what is needed when the team specifically wants high-value orders held for human review. Together, a targeted knowledge source and a deliberate autonomy threshold solve both the accuracy problem and the oversight requirement, while the other two options either remove needed automation or remove needed oversight.