Investigate fraud rings
on a single canvas.
ZIC Engine turns case data into an interactive relationship graph with AI-assisted triage, tamper-evident evidence, and a complete audit trail — so an analyst moves from raw data to a defensible decision in one place.
One screen, from first signal to final decision.
ZIC Engine presents the entities and transactions in a case as an interactive graph alongside an investigator workspace — no switching between tools to read a ring, check an entity, or record a decision. Upstream, real-time decisions come from the Zarelva Signal Engine, which hands high-risk events to ZIC as analyst-ready cases.
Import
Load entities and transactions from CSV or a saved case bundle. The data validates and renders straight into the graph.
Detect
Surface coordinated fraud rings and the cross-links between them from connectivity in the data.
Investigate
Spotlight a ring, drill into entities, resolve duplicates, and weigh an explainable AI recommendation.
Decide
Record a standardised outcome, backed by verified evidence and a complete audit trail.
Built for ring analysis, not just alerts.
Everything an analyst needs to work a multi-entity case from one workspace.
Visual investigation
See rings, entities, and money flows on one canvas; click any ring to isolate it.
Automated ring detection
Group connected parties into candidate rings and surface the links between them.
Entity resolution
Identify likely duplicate parties and merge them into a single canonical entity.
AI-assisted triage
A recommendation, model confidence, and the explainable risk factors behind a score.
Tamper-evident evidence
Hash attachments with SHA-256 and verify integrity on demand.
Complete audit trail
Every material action is recorded, so a case can be reconstructed from the record alone.
Four measures, kept distinct.
ZIC Engine never conflates severity with certainty. Each measure answers a different question.
Overall Case Risk
Case-level severity (0–100), driven by the highest-risk entity. A prioritisation signal — not a verdict.
Model Confidence
How certain the AI is in its recommendation. A measure of certainty — not of risk.
Ring detection confidence
Per ring, how strongly a cluster reads as coordinated. A prompt to review — not proof.
Risk factors
The itemised signals behind a score, so the reasoning stays transparent and reviewable.
Knowledge retrieval & investigation memory, built in.
ZIC doesn't just display a case — it remembers, correlates, and grounds every decision in evidence and regulation. No separate tools, no copy-pasting between systems.
Signal correlation
Signals from devices, transactions and identities are correlated across the case automatically — patterns surface without manual cross-referencing.
Investigation memory
Past cases, dispositions and confirmed patterns stay retrievable — a returning fraud ring is recognised, not re-investigated from scratch.
Evidence retrieval
Every piece of evidence in a case is indexed and retrievable in context, with SHA-256 integrity intact — findable at decision time, not buried in folders.
Regulatory knowledge retrieval
Decisions can be grounded against the applicable regulatory guidance — retrieved with provenance, never fabricated.
Explainable decision generation
The closing narrative is generated from the case's own signals, evidence and citations — a written rationale an auditor can follow line by line.
Evidence you can prove. Actions you can trace.
Two things make a case easy to defend in review.
SHA-256 evidence
Files are fingerprinted on attachment; verify integrity on demand to confirm nothing has changed.
Append-only audit
Imports, flags, merges, accept/override, and the final decision are all recorded automatically.
Standardised disposition
Every case closes with an outcome and a rationale, mapped to a clear decision framework.
A working demo — with honest boundaries.
We would rather you trust the tool than be dazzled by it.
- Synthetic data only. Every name, account, and transaction in the demo is fictitious.
- Evaluation vs production. In this demo, selected AI narrative and enrichment content is illustrative; a production deployment is configured against your own data, models, and intelligence sources.
- What is real. Ring detection, entity resolution, SHA-256 evidence verification, the audit trail, and case save / export work as shown.
- Analyst responsibility. Findings should be corroborated against authoritative systems before any regulatory filing or customer action.
Open the canvas.
Work a case end to end.
The live demo runs in your browser. Import the sample data, run ring detection, investigate the graph, verify a piece of evidence, and record a decision — the full investigation loop.
Launch the Live Demo →- Live, in-browser investigator canvas
- Sample fraud-ring case included
- Operating Guide (PDF) download
- Ring detection & entity resolution
- SHA-256 evidence verification
- Explainable AI recommendation
- Full audit trail & case export
- Built by an independent fraud practice