Skip to content
Agent packs

Lending · Agent pack

Credit decision model review

Cadre pack · Backstop

Tests a committed decision for policy consistency, defensibility, and unsupported overrides.

Demonstrated on representative cases. Source coverage, thresholds, policy interpretation, escalation design, and performance require institution-specific validation.

Operating dossier

The job, permitted decisions, and owner

Mandate

Challenges a committed credit decision against supplied policy and file evidence, and considers any fair-lending review on file, without re-underwriting or issuing a validation or legal opinion.

Case arrives with
Decision · Application · Decision · Product · Decision · Decision · Decision · Terms
Allowed outcomes
Pass · Findings · Fail
Where it fits
Application to Decision
Decision logic
  1. 01Second-line challenge of a committed credit decision: tests whether the decision is policy-consistent and defensible without pretending to be a full model validation.
  2. 02Checks consistency against policy and against similarly-situated files decided before — on credit merit only, never protected attributes.
  3. 03Tests whether the stated factors genuinely support the outcome; a decline resting on factors that do not fail policy is a finding.
  4. 04Treats deviations from the supplied policy as overrides needing documented justification; an observed decision can receive an internal pass result without implying model validation.
  5. 05Returns pass, findings, or fail as a case-level rating, naming the comparator or factor behind anything below pass.
Must not
Use protected-basis fields in the decision path.
Human owner
Model-risk findings and remediation
Evidence retained
Structured outcome, observed path, and the source evidence used.

Case and record

One case, from supplied evidence to recorded decision

Supplied case

Representative case dataSynthetic or simulated data
Decision · Application
L1
Decision · Product
auto
Decision · Decision
approve
Decision · Terms
3 fields
Decision · Score band
prime
Decision · Dti pct
11
Decision · Ltv pct
103
Decision · Policy basis
score 768 in the prime band, DTI within the maximum, LTV within the maximum
Decision · Factors
strong credit score, low revolving utilization, clean payment history
Decision · Confidence
0.96
View the case data

Case key L1-clean

{
  "decision": {
    "applicationId": "L1",
    "product": "auto",
    "decision": "approve",
    "terms": {
      "apr": 6.49,
      "amount": 31000,
      "termMonths": 60
    },
    "scoreBand": "prime",
    "dtiPct": 11,
    "ltvPct": 103,
    "policyBasis": [
      "score 768 in the prime band",
      "DTI within the maximum",
      "LTV within the maximum"
    ],
    "factors": [
      "strong credit score",
      "low revolving utilization",
      "clean payment history"
    ],
    "confidence": 0.96
  }
}

Evaluation and limits

Evaluation coverage and stopping points

Published evaluation map

Expected decisions, boundaries, and costly failure modes.

These cases show what the pack is asked to decide, when it should stop, and which plausible errors the evaluation is meant to expose. They are not a reliability score or independent validation.

Expected and borderline behavior

Routine judgments, close calls, and named stopping points.

  • prime approve, within policy, factors support it

    pass

    Case key

    L1-clean
  • decline rests on hard policy fails (score below min, LTV over max)

    sound; pass, minor findings acceptable

    Case key

    L2-supported
  • within-policy near-prime file declined with no hard basis; similarly-situated L12 was approved

    findings/fail

    Case key

    L11-inconsistent

Operating fit

Workflow placement and validation

Operating pattern

Independent review

Separate the work from the party that tests it, then require findings to point back to evidence and a named standard.

Decision sequence and workflow placement
  1. 01Receive completed work
  2. 02Reconstruct the standard
  3. 03Run independent checks
  4. 04State findings and limits
  5. 05Assign accountable review
Pattern guide

Institution-specific validation

These questions remain open until the pack is fitted to an institution.

Validation questions
  • Policy and legal interpretation
  • Source coverage, quality, and freshness
  • Thresholds and exception calibration
  • Integration and degraded-mode behavior
  • Human approval and escalation design
  • Performance on the institution’s own case mix
Inputs, actions, and outcomes
Give it
  • Decision · Application
  • Decision · Product
  • Decision · Decision
  • Decision · Terms
  • Decision · Score band
  • Decision · Dti pct
  • Decision · Ltv pct
  • Decision · Policy basis
  • Decision · Factors
  • Decision · Confidence
It does
  • Get applicationget_application
  • Get credit reportget_credit_report
  • Get policyget_policy
  • Find similarfind_similar
  • Record assessmentcommitrecord_assessment
It returns · Rating
  • Pass
  • Findings
  • Fail
Related packs

Related packs