Car Finance Utility — AutoFab

Version: 1.0 Last Updated: May 2026 Reference Implementation: AutoFab

Component Overview

The Car Finance utility is a motor finance remediation application built on the DataFab core framework. It supports the end-to-end remediation lifecycle of motor finance claims under the FCA Motor Finance Scheme — from data ingestion and population building, through cohorting, scheme testing, redress calculation, customer outreach, payment validation, and case closure with a complete audit trail.

The flagship deployment of this utility is AutoFab — a standalone instance in a private cloud that supports execution services. The Car Finance utility is structured so that the same capabilities can be reused for other motor-finance remediation engagements with different banks, schemes, and integration points.

Capabilities:

Architecture Overview

┌──────────────────────────────────────────────────────────────────────┐
│                CAR FINANCE UTILITY (AutoFab Reference)               │
├──────────────────────────────────────────────────────────────────────┤
│                                                                      │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │                    DATA SOURCES (Bank-side)                    │  │
│  │  Customer Records │ Loan Agreements │ Payment Histories        │  │
│  │  Vehicle Data │ Dealer Data                                    │  │
│  │  Connected via VPN tunnel / Private Link                       │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │  Stage 1   KNOWLEDGE FABRIC                                    │  │
│  │            MCP Connectors → Discovery → Schema Mapping →       │  │
│  │            Data Quality Validation                             │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │  Stage 2   POPULATION BUILD                                    │  │
│  │            Entity Resolution → Knowledge Graph                 │  │
│  │            Customer → Loan → Vehicle → Dealer relationships    │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │  Stage 3   COHORTING (Graph Operations)                        │  │
│  │            Rule Engine → Cohort Assignment → Document Store    │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │  Stage 4   SCHEME TESTING (Studio DDA Pipelines)               │  │
│  │            Deterministic eligibility tests → Calculated flags  │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │  Stage 5   REDRESS CALCULATION (Studio DDA Pipelines + HITL)   │  │
│  │            Scheme formulae → Calculated amounts → Audit trail  │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │  Stage 6   STP / PORTAL / OUTREACH (Chain of Agents)           │  │
│  │            Auto-progress clear cases │ Amazon Connect outreach │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │  Stage 7   PAYMENT VALIDATION (Studio DDA Pipelines)           │  │
│  │            Validate completeness → Generate payment file       │  │
│  │            Encrypted SFTP delivery to bank                     │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │  Stage 8   CLOSURE & AUDIT (Graph Operations + Knowledge       │  │
│  │            Fabric Write-back)                                  │  │
│  │            Close case │ Update bank CRM │ Compile audit trail  │  │
│  └────────────────────────────────────────────────────────────────┘  │
│                              │                                       │
│  ┌────────────────────────────────────────────────────────────────┐  │
│  │           ANALYST DIALOG (Playbook)                            │  │
│  │  list_cases │ explain_decision │ rule_deep_dive │ outreach     │  │
│  │  payment_status │ audit_trail                                  │  │
│  └────────────────────────────────────────────────────────────────┘  │
└──────────────────────────────────────────────────────────────────────┘

The utility uses Studio’s Chain of Agents rather than Graph Operations DAG Rules as its primary orchestration: the remediation lifecycle is naturally a sequence of stages with explicit Human-in-the-Loop gates (e.g., redress calculation approval) and integration boundaries (Amazon Connect, payment delivery, CRM write-back). The Graph Operations Rule Engine is used inside Stage 3 (Cohorting) for cohort assignment scoring.

Domain and Schemas

Schemas are managed by the Schema Registry. All extraction is schema-bounded — the platform can only extract and process attributes defined in the Car Finance domain schema; bulk replication of bank data is not permitted.

Data Sources

Connection security (see 02-Architecture: Connectivity, 03-Knowledge-Fabric: MCP Connectors):

Eight-Stage Remediation Lifecycle

The Car Finance utility implements an eight-stage remediation lifecycle. Each stage maps to specific DataFab core components.

Stage 1 — Data Source & Cleanse

Stage 2 — Population Build

Stage 3 — Cohorting

The cohorting Rule Engine uses the Pattern Scoring Rules defined in 08-Graph-Operations. Rule changes follow Rule Engine governance (role-based editing, version control, multi-party approval, testing, rollback).

Stage 4 — Scheme Testing

DDA pipelines in this stage are deterministic — they do not use probabilistic LLM-driven reasoning for the eligibility decision itself. LLM use is restricted to pre/post-processing tasks (e.g., document understanding) and bounded by schema validation and ontology execution per 05-AI-LLM.

Stage 5 — Redress Calculation

A Human-in-the-Loop gate is enforced on the redress calculation stage by default. Calculation parameters, intermediate values, and final amounts are persisted with the calculation audit trail.

Stage 6 — STP / Portal / Outreach

Amazon Connect data handling (see AutoFab Document 02, Section 12):

Stage 7 — Payment Validation

Payment-file delivery security:

Stage 8 — Closure & Audit Trail

Integration Points

The utility has four defined integration points with external systems.

Internal Data Store

Claim processing data is persisted in the internal Document Store.

Decision Outputs

Decisions are deterministic and reproducible: every decision carries a pointer to the rule version, DDA pipeline version, inputs, intermediate values, and approver (where HITL is engaged).

Dialog Playbook

The Car Finance playbook (see Dialog › Playbook-Based Communication) governs the analyst conversation surface. It is the analyst-facing layer over the case-management workflow.

Intents

Tool Catalog (illustrative)

Tools are subject to Tool Authorization; analysts only see tools their role permits.

Persona and Guardrails

System Self-Awareness

Within Car Finance, Self-Awareness (see Dialog › System Self-Awareness) gives analysts these specific capabilities:

Self-Awareness is read-only; rule, threshold, and pipeline changes go through the Rule Engine and DDA-publish governance workflows.

User Roles

The utility implements role-based access control aligned with the engagement structure (matching the AutoFab reference deployment).

Authentication: MFA required; password policy minimum 12 characters; idle timeout 15 minutes; absolute session timeout 8 hours; SAML 2.0 SSO supported.

Operational Modes

Given the regulatory nature of remediation, the Car Finance utility runs primarily under Mode 1 (AI-Assisted Manual) and Mode 2 (Routine Automation).

Modes 0 and 4 are not used for this utility; redress calculations always retain a HITL gate by default to meet regulatory expectations on fair treatment and decision rationale.

Reference Implementation: AutoFab

Deployment: Standalone single-tenant instance in a private cloud, with limited integration points. Scope: End-to-end remediation lifecycle (Stages 1–8) for motor finance claims.

Vendor (DataFab) access controls in the AutoFab deployment:

Security Controls

Audit Logging

Logs are stored in immutable tamper-evident storage with PII redaction. Auditor role has read-only access to all cases and audit logs.

Data Subject Rights

Regulatory Mapping

Recovery Objectives (Reference Implementation)

DR testing schedule: monthly backup-restore tests; quarterly failover tests; annual full DR exercise; semi-annual tabletop exercise.

Cross-References

• Overview — Utilities overview
• Transaction Monitoring — Reference utility implementation
• Compliance — Compliance utility (skeleton)
• Knowledge Fabric — MCP Connectors, Discovery, Entity Resolution, Knowledge Graph
• Studio — DDA Pipelines, Chain of Agents, Tool Authorization, Widget Types
• Dialog — Playbooks, Context-Awareness, System Self-Awareness, Widget Rendering
• Graph Operations — Rule Engine, Cohort Analysis, Case Management
• Compliance Capabilities — Regulatory framework, retention