PTI and Risk Engines
Risk engines (decision engines, credit policy platforms, rules orchestrators) evaluate institution-specific policies against applicant data, bureau scores, internal behavior, product rules, and regulatory constraints. PTI feeds risk engines with portable trust intelligence; it does not replace the institution's authority to approve or decline.
1. What risk engines are
Risk engines are policy execution platforms that combine data sources, models, and business rules into a decision outcome. Common forms:
- Credit decision engines: scorecards, cutoffs, reason codes
- Business rules management (BRM). Drools, FICO Blaze, custom rule DSLs
- ML model serving: real-time feature vectors and model endpoints
- Workflow orchestration: refer, manual review, stipulation paths
- Champion/challenger. A/B policy testing and monitoring
Risk engines answer: Given our policy and this applicant's data, what action do we take?
2. What problem risk engines solve
| Problem | Risk engine response |
|---|---|
| Inconsistent manual underwriting | Codified policy rules |
| Regulatory reason code requirements | Structured decline explanations |
| Product-specific cutoffs | Tiered pricing and limits |
| Model governance | Versioned scorecards and audit |
Risk engines are institution-owned decision systems. They require quality inputs: but typically integrate each data source through bespoke connectors with no standard trust intelligence envelope across vendors and contexts.
3. What PTI adds
Risk engines
- Institution policy execution
- Final approve / decline / refer
- Product-specific scorecards
PTI adds
- Standard trust input. JSON envelope with drivers and confidence bands
- Context-scoped features, request
lendingonly, not global mashup - Provenance metadata, feature sources for model documentation
- Coverage gaps, explicit thin-data flags for policy branching
PTI implements the Trust Intelligence Engine conformance class, producing inputs suitable for risk engine feature pipelines. The institution retains decision accountability; PTI supplies governed external intelligence.
4. How they compose together
Integration pattern:
- Origination workflow triggers parallel data fetches, bureau, internal, PTI trust lookup (
contexts: ["lending", "risk_compliance"]). - PTI response maps to feature variables.
pti_lending_confidence_pct,pti_driver_repayment_weight,pti_coverage_gap_count. - Risk engine applies institution scorecard. PTI features are one input block: not the decision itself.
- Decline reason codes combine bureau reasons with PTI driver labels where policy requires.
Institutions SHOULD treat PTI as adjacent data subject to fair lending and model risk management, documenting feature usage in model cards.
5. When to use each
| Scenario | Risk engine | PTI |
|---|---|---|
| Final loan approve/decline | Required | Input only |
| Portable cross-MFI behavioral features | Engine cannot source alone | Required |
| Internal overdraft limit rules | Required | Optional |
| Multi-context tenant + employment check | Engine needs features | PTI multi-context lookup |
| Regulatory decision ownership | Institution | PTI never decides |
Rule of thumb: PTI is upstream intelligence; the risk engine is downstream policy.
6. Related PTI spec/RFC links
- RFC-004. Trust Lookup API
- Explainability guide (
explain_score.v1) - RFC-012. Trust Evidence
- Reference API Specification
- Compliance guide