Enterprise knowledge systems, legal profile models, and review-forecasting tools all depend on evidence that is relevant, scoped, chronological, and auditable. Dense retrieval and large document encoders improve semantic recall, but governance-sensitive workflows also require entity boundaries, access controls, temporal validity, citation provenance, and calibrated uncertainty. This paper surveys enterprise knowledge, legal profiles, and review forecasting as one evidence-lifecycle problem. We synthesize work on governed cloud pipelines, cross-cloud workload control, multi-cluster gateways, hybrid semantic-relational retrieval, long-horizon forecasting, historical legal profiles, reviewer agreement profiles, agreement-gated self-ensembling, structured extraction, and GPU-parallel optimization. The survey identifies five recurring design tasks: evidence acquisition, chronology-safe profile construction, constraint-aware retrieval, confidence-gated prediction, and audit-preserving publication. A comparative coding of representative systems shows that retrieval quality and predictive discrimination are most useful when paired with explicit validity checks and evidence-pack provenance. The conclusion is pragmatic: enterprise and review-support systems should treat evidence construction as a first-class control plane, not as a pre-processing detail.
Mazumder et al. (Fri,) studied this question.