Skip to main content
REAPS
Governance Framework

The REAPS Framework

Responsible, Explainable, Auditable, Policy-Driven, Sovereign

Enterprise AI governance ensuring every decision is ethical, transparent, traceable, compliant, and under your control. Enforced at the infrastructure layer, not bolted on after the fact.

The Five Principles

Each principle is enforced at runtime across every AI interaction, ensuring enterprise-grade governance without compromising performance.

Infrastructure Enforcement

L0-L3 Policy Hierarchy

REAPS is not a checklist -- it is enforced at the infrastructure layer through a strict 4-level policy hierarchy. Lower layers can tighten but never loosen upper-layer constraints.

L0

Framework

All tenants, all products

Immutable REAPS constraints. Cannot be overridden by any layer. Defines the absolute boundaries for all AI behavior.

AI must never fabricate data sources.
All decisions must be reproducible.
Human escalation required above confidence threshold.
L1

Domain

All products in a domain

Industry-specific policies for BFSI, healthcare, and other regulated verticals. Inherits L0 constraints.

BFSI: Credit decisions require dual-model validation.
Healthcare: PHI data restricted to HIPAA-compliant endpoints.
Payments: ISO 20022 message validation mandatory.
L2

Product

Single product, all tenants

Product-specific policies that define operational boundaries. Each Clarvus product has tailored L2 policies.

CV-STANLI: Forecasts must cite minimum 3 data points.
CV-Certifier: Assessment timeout at 30 minutes.
arthaLOAN: Risk score must be explainable in 5 factors or fewer.
L3

Tenant

Single tenant

Organization-specific configurations. Can tighten but never loosen higher-layer policies. Maximum flexibility within guardrails.

Bank X: Require CFO sign-off for amounts > 50L.
NBFC Y: Restrict AI usage to business hours only.
FinTech Z: Enable experimental features with sandbox isolation.
Each layer inherits and can only tighten constraints from the layer above.
Governance Mechanism

Human-in-the-Loop (HITL)

AI augments human decision-making without replacing accountability. Every critical decision maintains human oversight with confidence-aware approvals.

AI proposes but does not conclude

AI generates recommendations and analysis but final decisions rest with human experts who understand context and consequences.

AI explains but does not obscure

Every AI output includes clear reasoning that domain experts can validate. No black-box decisions in regulated environments.

AI learns but never invisibly

Model improvements and adaptations are logged, reviewed, and approved. No silent updates that change system behavior.

Accept / Modify / Reject Cycle

1

Accept

Human approves AI recommendation with full confidence.

2

Modify

Human adjusts AI output with tracked changes.

3

Reject

Human overrides AI with documented reasoning.

Confidence-Aware Approvals

High
Auto-approve with audit
Medium
Human review required
Low
Escalation mandatory

REAPS Compliance Levels

Choose the governance level that matches your regulatory requirements and risk profile.

Basic

RA

Internal tools and development

Standard

REAP

Customer-facing applications

Enterprise

REAPS

Regulated industries

Critical

Enhanced
REAPS+

Financial decisions, healthcare

Executive Engagement

Ready to Implement REAPS Governance?

Schedule a tailored session with our solutions team. We will walk you through how REAPS can be implemented for your specific regulatory requirements and risk profile.