A consultant reviews an opportunity in Dynamics 365 Sales for a manufacturing client where the AI-generated estimated deal value is $40,000, but the seller's actual quote, built in a separate CPQ system, totals $95,000 once product bundles and volume discounts are applied. The discrepancy is large enough that sales leadership no longer trusts the AI-generated value on any opportunity. What is the most likely cause and correct remedy?
Select an answer to reveal the explanation.
Short Explanation
The estimate isn't wrong because the AI is bad at math — it's wrong because it's doing math on incomplete information. If the real deal, bundles, discounts, and all, only exists inside a separate quoting tool and never makes it back into the opportunity's line items, the estimate is built from whatever's left on the CRM side, which in this case is a fraction of the real number. That's a plumbing problem: get the quote data flowing back into the opportunity record, and the estimate catches up automatically. Retraining a model for one client's deal sizes isn't how this works, and writing off the whole feature because one integration gap produced a bad number throws out something that's accurate everywhere the data actually is complete. Having a seller manually patch the number on every single opportunity might paper over today's gap, but it's not a fix — it's just extra work that has to happen forever.
Full Explanation
The correct answer is A. The AI-generated estimated deal value is calculated from the opportunity's product line items as recorded in Dataverse, so if the bundled products and volume discounts that make up the seller's actual $95,000 quote were built and finalized in a separate CPQ system without being synced back into the opportunity, the estimate will reflect only the partial or list-price data still on the record, producing the large gap leadership is seeing. Syncing the CPQ output back into the opportunity's line items closes that gap. Option B is incorrect because the estimation model does not require per-client manual retraining to account for a given deal's pricing; the issue here is missing input data, not an untrained model. Option C is incorrect because concluding the feature only works for simple, non-discounted deals overgeneralizes from one data-sync failure and discards a capability that works correctly once line items are complete. Option D is incorrect because having sellers manually overwrite the estimated value field on every opportunity is not sustainable across a full pipeline and does not fix the underlying integration gap causing the mismatch.