A sales director notices that the Dynamics 365 Sales AI-predicted forecast for a business unit totals $1.2 million higher than the sum of what individual sellers marked as their committed forecast category for the same period. Several sellers say their committed numbers are accurate and reflect deals they are confident will close. The director asks a consultant how to reconcile this without simply telling sellers to raise their commit numbers. What should the consultant recommend?
Select an answer to reveal the explanation.
Short Explanation
When the machine's number and the human's number don't match, the instinct is to pick a side — trust the algorithm or trust the seller. Neither is the move. The useful question is what the model is actually seeing: maybe it's picking up on email response patterns or a resemblance to deals that closed quickly last quarter, things a busy seller hasn't consciously logged yet. Pull on that thread deal by deal and you'll usually find either a real blind spot in the human's read or a pattern the model is leaning on too hard for this particular account. What you don't want to do is quietly turn off the predictive layer because it's inconvenient, and you definitely don't want to strong-arm sellers into padding their commit numbers just to make the totals line up — that just trades an honest disagreement for a false one. Same goes for hand-editing probability fields until the math works; that's cosmetic, not diagnostic, and next quarter you'll be right back here.
Full Explanation
The correct answer is C. When an AI-predicted forecast diverges sharply from seller-committed numbers, the productive path is to examine which specific opportunities are driving the gap and what signals the model is weighing on those deals, then compare that against what sellers actually know about the account. This surfaces whether the model is catching real momentum sellers have not yet logged, or whether it is overweighting patterns that don't apply here, letting the director act on evidence rather than gut feel. Option A is incorrect because disabling AI-predicted forecasting discards a signal that may be catching genuine deal movement sellers haven't recorded, and does nothing to explain the current gap. Option B is incorrect because pressuring sellers to inflate their committed numbers to match the AI figure risks creating inaccurate commitments that will look worse when deals slip, rather than resolving the underlying disagreement. Option D is incorrect because manually adjusting close probability fields to force alignment manipulates the data rather than investigating why the model and the sellers disagree, and provides no lasting reconciliation process.