Fraude Intelligence

The Fraude Intelligence example shows Scrydon set up for a financial-services AML / fraud-detection analyst. It ties together managed tables, the typed ontology layer, and an agentic workflow — so the analyst can ask plain-language questions and get governed, typed, cited answers.

What it models

A regulator's view of:

• Transactions between counterparties.
• Customers (the persons / entities party to those transactions).
• Suspicious activity reports (SARs).
• Regulated entities and the regulators that supervise them.
• Risk assessments and supervisory actions.

The example ships a Fraude Intelligence ontology pack with pre-built bindings against the demo CSVs. Once installed, the analyst can ask:

• "Show me all transactions involving Acme Holdings that triggered a SAR in Q1."
• "Which regulators supervise the high-risk entities in Belgium?"
• "What's the trend in supervisory actions against fintechs?"

Download

The pack is a downloadable .scrydon-extension.tar.gz bundle — it is not preshipped with the platform.

• fraude-intelligence.scrydon-extension.tar.gz — latest.
• scrydon-pack-fraude-intelligence-1.0.0.scrydon-extension.tar.gz — version-pinned.

Install

1. Download the bundle above.
2. In your Scrydon deployment, open Settings → Platform → Extensions → Sources and click Upload a pack (one-off).
3. Drag the .tar.gz onto the drop zone (or click to pick the file). The dialog shows the bundle's packageId, version, and ontology contributes summary.
4. Tick the unsigned-pack acknowledgement and click Upload. The pack appears in your org catalog as Fraude Intelligence v1.0.0.
5. In your workspace, open Analytics → Marketplace and install Fraude Intelligence for the active workspace-environment.
6. Upload the demo CSVs below as managed tables, then map the pack's bindings to them in /ontology → Bindings.

For the full walkthrough (schema inspection, binding column maps, graph verification, building an agentic triage workflow), see the step-by-step tutorial.

What's interesting about it

• Typed reasoning — the agent reasons about RegulatedEntity and Transaction, not about regulated_entities rows. Renaming a column doesn't break the workflow.
• Provenance everywhere — every agent answer cites the underlying rows and bindings.
• Multi-hop traversal — ask about regulators of entities involved in high-risk transactions; the agent traverses the links without you wiring the joins.
• Column-level masking — confidential fields (e.g. originatorName) are masked for members and visible to admins.

Sample data

Nine anonymised, cross-referenced CSVs in apps/docs/public/static/fraude-intelligence/:

• compliance_authorities.csv
• customer_due_diligence.csv
• persons.csv
• regulated_entities.csv
• risk_assessments.csv
• supervisory_actions.csv
• supervisory_authorities.csv
• suspicious_activity_reports.csv
• transactions.csv

Related

• Step-by-step tutorial — install, bind, verify, build an agentic workflow.
• Ontology — the typed layer.
• Ontology → Packs — how packs ship and install.
• Analytics → Classification & masking — column-level governance.
• Compliance → GDPR — applies for personal data in this scenario.