Evaluator Context · Recommendation Layer

POLICY — Recommendation

The policy stage converts the AI trust assessment and related explanation context into an advisory recommendation such as APPROVE, REVIEW, STEP-UP AUTHENTICATION or REJECT. This is a governed layer distinct from the AI model itself.

POLICY — Recommendation
Expanded evaluator context
Professional interpretation

Purpose in the demonstrator

This stage demonstrates institutional control. The AI does not authorize payments. Policy interprets AI outputs according to defined thresholds and decision logic, while the participating institution retains final authority.

Operational meaning

Separation of concernsAI assessment and policy recommendation are intentionally distinct.
Governed logicRecommendations come from explicit decision rules, not unexplained model behaviour.
Institutional boundaryThe portal reminds evaluators that recommendation is advisory, not final transaction authorization.

Inputs and outputs

ElementMeaning in this demonstrator
AI trust outputsTrust score, risk category and confidence entering the policy engine.
Policy rule setVersioned logic that turns assessment into an advisory operational recommendation.
RecommendationAPPROVE, REVIEW, STEP-UP AUTHENTICATION or REJECT.
OutputGoverned institution-facing advisory result.

Evaluator questions this page answers

  • How is recommendation different from model scoring?
  • Where is institutional control preserved?
  • Can the same trust outcome lead to different recommendations under different policy thresholds?
  • Why is policy an essential governance layer?

How this stage fits in the end-to-end chain

1. AI TRUSTAssessment available.
2. EXPLAINReasons increase interpretability.
3. POLICYControlled recommendation generated.
4. AUDITRecommendation and rationale are retained for traceability.

Governance and interpretation notes

This page is an evaluator-facing contextual explanation of the demonstrator stage. It explains the business role, the technical contribution, the governance boundary and the expected interpretation of the displayed output. It is designed to make the demonstration credible, understandable and professionally reviewable without exposing raw internal JSON on-screen.

Policy Decision Boundary

This graphic separates AI assessment from recommendation logic and preserves the institution’s final decision authority.

Policy Decision Boundary
Policy converts assessment context into APPROVE, REVIEW, STEP-UP AUTHENTICATION or REJECT while keeping final authorization external.