Trace
Compliance needs a causal record, not a screenshot of the final score.
A control layer for automated loan decisions: trace every fact, block invalid influence, replay appeals.
Applicants get a conclusion. Compliance needs evidence. The gap is the black box.
Compliance needs a causal record, not a screenshot of the final score.
Applicants need to know which fact can actually change the outcome.
The best compliance review happens before the denial leaves the system.
Trained on a 150,000-record lending dataset, so the scoring behaviour Glass Box traces reflects real applicant patterns instead of a toy heuristic.
A second model trained specifically to understand a decision: it reads the full factor trace, checks which influences were valid, and flags what a reviewer needs to see.
The model denies her. Glass Box reconstructs the decision path before the denial reaches Maya.
Only the affected branch is replayed. Debt-to-income improves and the outcome moves to review.
See the facts that mattered and the evidence worth submitting.
Catch prohibited influence before a decision is released.
Keep automated decisions fast without losing a defensible record.