A sales enablement consultant is rolling out an AI-generated lead scoring model. During a pilot, sellers notice the model consistently assigns high scores to leads from a specific industry segment that historically converted well, but the company is now expanding into new industries where it has little historical data. Sellers worry the AI will systematically undervalue promising leads from these new segments. What should the consultant recommend to address this concern?
Select an answer to reveal the explanation.
Short Explanation
When a model has barely seen a new kind of lead before, its confidence is really just a guess dressed up as a number, and treating that guess as gospel is where the risk comes in. The answer is not to rip the feature out for that segment, and it is definitely not to sit on your hands for a year waiting for the data to show up on its own. It is also not safe to just borrow a score from some other segment that seems kind of similar, because kind of similar can hide a lot of important differences. What actually works is letting the score coexist with rules a person can set and adjust, so sellers can lean on their own judgment for the segments where the AI does not have enough history yet. As real conversions start coming in from the new segment, the model gets smarter and the manual weighting can gradually step back.
Full Explanation
The correct answer is B. When a model has thin or no historical data for a new segment, its scores for that segment carry higher uncertainty, and the practical mitigation is to let sellers apply configurable criteria or override rules alongside the AI score so human judgment can fill the gap until enough conversion data accumulates for the model to learn the new pattern. Option A is incorrect because removing the score entirely leaves sellers with no prioritization signal at all, which is a step backward rather than a fix, and abandons a capability that still adds value for existing segments. Option C is incorrect because pausing evaluation for a full year means sales activity on those leads proceeds without any structured prioritization in the interim, which is impractical and unnecessary when a blended approach is available immediately. Option D is incorrect because applying a score from a superficially similar old segment assumes the segments behave alike, which is precisely the unverified assumption that could mislead sellers and is riskier than acknowledging the uncertainty directly. Blending AI scoring with adjustable manual weighting gives sellers a usable signal now while the model catches up on new-segment data over time.