Evidence before conclusions

AI reads supported files and returns structured candidate facts with short page-linked excerpts and confidence. It does not set the scores, risk level, market range, final-cost scenario, comparison order, or recommendation.

A consequential document fact needs qualifying document evidence before deterministic rules may use it. Missing, contradictory, or low-confidence information stays unresolved. Low extraction confidence blocks the affected finding instead of displaying a false score.

Versioned application rules calculate clarity, fine-print risk, known-cost normalization, and realistic scenarios. ScopeLock uses supported service and inventory facts only. ChangeWatch compares supported versions uploaded together; it is not continuous monitoring.

The production market benchmark source set is currently inactive. Until a compatible, dated, methodologically supportable price distribution is admitted, the report shows “Not enough information to score” instead of presenting a market range or Price Fairness Score.

Documents can be incomplete, ambiguous, inconsistent, or wrong. Extraction can miss or misclassify text. A scenario is not a guarantee, and a flag is not a legal conclusion. Use the evidence links to check the original language and confirm important details with the mover.