Skip to content

Case Management

Documentation status

Last verified: 26 Sep 2026. Readiness uses the single rubric and status vocabulary.

Two different things share the word “case” in MasteryHive conversations. This page separates them so they cannot be confused.

Concept What it is Product page boundary
AI case-rule evaluation Synchronous rule scoring for case-oriented decisioning Coded in Fraud AI products
Durable case management Full investigator lifecycle (queue, assignment, notes, approval, closure, …) Not evidenced in this documentation’s source repositories

Purpose

  • Document what the case-management rule engine actually does.
  • State clearly which durable case-desk capabilities are unverified because the banking dashboard / application backend that would own them is not present in the MasteryHiveAI documentation source tree.

Users

Audience Relevance
Partner engineers May call the case-rule scoring API
Fraud / ops stakeholders Must not assume a full case desk from AI scoring alone
Product evaluators Distinguish rule evaluation from investigator workflow

Part A — AI case-rule evaluation (coded)

Purpose

Evaluate caller-supplied rules against one transaction and return a prediction immediately — isolated so bursts of this traffic do not contend with the main rule scoring path.

Capabilities (confirmed)

Capability Status
Synchronous rule evaluation IMPLEMENTED
Authenticated scoring (Authorization + tenant header) IMPLEMENTED
Prediction with flagged / score / condition lists IMPLEMENTED
Durable case record (case_id, queue, assignee) Not in this service
Report publish or post-action from this service Not in this service

End-to-end workflow

flowchart LR
  caller[Banking_backend]
  rules[Case_rule_engine]
  pred[Prediction_response]
  caller --> rules --> pred
  1. Backend sends a transaction event + rules.
  2. Engine evaluates conditions and returns a prediction.
  3. Caller decides what to do next (including whether to open a durable case elsewhere).

Inputs / outputs

Inputs Outputs
Transaction / user context + rules list Prediction: flagged, score, passed/failed conditions

Partner path: POST /case_management_rule_engine — see Fraud API.

Integrations

  • Banking application backend (caller)
  • Same Fraud auth pattern as other scoring routes — Auth and tenancy

Readiness (rule engine only)

Rated with the single readiness rubric. Reasons for each rating are on that page.

Gate Rating
Code PARTIAL (development branch only)
Integration Verified UNKNOWN
Ops Ready PARTIAL
Security Approved NOT MET
GA NOT MET

Part B — Durable case management (dashboard / backend)

Evidence status

Capability requested for product docs Evidence in MasteryHiveAI AI / docs source tree
Approval paths (case approve / reject workflow) UNKNOWN — not found
Case types taxonomy UNKNOWN — not found
Case levels (desk levels / ownership tiers) UNKNOWN as case levels (scoring has risk labels only)
Post-actions after scoring PARTIAL — Fraud services can publish a post-action trigger; no consumer is verified in this tree
Maker-checker UNKNOWN for TM cases
Assignment UNKNOWN
SLA escalation UNKNOWN (report may carry escalation labels only)
Evidence on the case PARTIAL — evidence exists on the fraud report payload, not as a case workspace
Notes UNKNOWN for TM cases
Approval / closure / reopening UNKNOWN for TM cases
Audit history UNKNOWN for TM cases

Do not treat durable case management as implemented

Searches of the Fraud, KYC, and related AI repositories found no durable TM case entity, queue, assignment, notes, approval, closure, reopen, or SLA workflow. Product notes and screenshots outside this tree are not accepted as implementation evidence here. If a case desk exists, it lives in the banking application backend / dashboard — behaviour UNKNOWN until that codebase is documented from source.

Banking Reconciliation has an exception resolve lifecycle (approve / reject / escalate / reopen / reconcile) with append-only audit. That is finance exception management, not fraud case management — see Reconciliation.

Fraud report handoff (not a case desk)

Fraud report compose delivers decision, escalations, evidence, and narrative to the banking application’s create-report endpoint. That creates an investigator-facing report, not a verified full case lifecycle.

Business problem (why the confusion exists)

Stakeholders often say “case management” for both (1) scoring a transaction with case-style rules and (2) running an investigator desk. MasteryHive AI products implement (1). Claiming (2) without dashboard source evidence would be inaccurate.

Readiness summary

Product slice Code Integration Verified Ops Ready Security Approved GA
Case-rule evaluation API PARTIAL UNKNOWN PARTIAL NOT MET NOT MET
Durable TM case desk UNKNOWN UNKNOWN UNKNOWN UNKNOWN NOT MET

Report evidence and escalation labels are IMPLEMENTED on Fraud reports and rated with Fraud.

See Product readiness summary.

Limitations

  • The service name includes “case management” but implements rule evaluation only.
  • Do not map maker-checker, SLA, assignment, or closure to the AI rule engine.
  • Do not invent dashboard behaviour from meetings or screenshots.