Agentic SAR drafting
Multi-agent workflow drafts SARs in NCA-compliant structure from your case material. Investigator reviews, edits, and approves — the human stays the accountable signatory.
The world's most defensible fraud-tracing platform
Trace financial crime end to end. Draft regulator-grade SARs in minutes.
VigilPath is the agentic investigation workbench for regulated financial institutions. Mule-network graphing, sanctions and PEP intelligence, behavioural and device signals, crypto-fiat bridging, and explainable AI — all wrapped around an investigator-in-the-loop review workflow with a full immutable audit trail.
- Cases handled
- 4,000+
- per investigator / year at Tier-1 banks
- SARs drafted
- 1,000+
- per investigator / year
- Time per SAR
- < 8 min
- with VigilPath, versus 45 m–3 h
Product
A complete investigation surface — not a feature bolted onto a TM system.
VigilPath sits one step downstream of your transaction-monitoring system and replaces the fragmented spreadsheet-and-Word-document workflow that consumes most of an investigator's day. Every capability below is in the platform.
Live module
Agentic SAR drafting
Watch the agent compose a regulator-ready narrative section by section.
Drafting next section…
Every capability, built into the platform
Twelve modules. One canonical case record. One immutable audit trail.
Regulator-usability scoring
Every draft is scored against an internal usability rubric (factual grounding, typology alignment, regulatory completeness, clarity, defensibility) with specific rewrite suggestions.
Mule-network graphing
Cross-case graph surfaces shared devices, counterparties, and transaction rings — turning isolated alerts into the network they actually belong to.
Cross-border typology library
Curated typology library covering high-risk corridors, trade-based laundering, shell-corporate layering, structuring and threshold evasion — versioned and citable inside every narrative.
Sanctions & PEP intelligence
Continuous screening against OFSI, OFAC, EU and UN lists with second-degree UBO resolution and adverse-media synthesis surfaced inline on the case.
Crypto-fiat bridge detection
On-ramp / off-ramp tracing across major centralised exchanges and stablecoin rails, with attribution clusters and exposure scoring carried into the SAR narrative.
Behavioural & device signals
Session, device, and behavioural-biometric signals from your fraud stack are correlated against investigation context to lift weak alerts into actionable cases.
Real-time risk scoring
Continuously re-scores in-flight cases as new evidence arrives — so MLROs see priority shifts the moment they happen, not at the next batch run.
Federated intelligence (opt-in)
Privacy-preserving, hashed indicator sharing between consenting tenants. No PII leaves your boundary — only typed indicators of compromise.
Explainable AI (XAI)
Every typology call, score and suggestion ships with its evidentiary basis. Reviewers see exactly which facts drove the conclusion — no black box, audit-ready.
Multi-jurisdiction SAR export
Structured export to NCA SAR Online (UK), FinCEN (US), FINTRAC (CA) and AUSTRAC (AU) schemas from the same canonical case record.
Regulatory horizon scanner
Tracks JMLSG, FCA, FATF and HM Treasury updates and flags cases where a guidance change alters the disposition rationale.
Architecture
Six-layer agentic pipeline. One accountable human.
The platform separates ingestion, context, typology, narrative, quality and review into independently testable layers. Each layer is observable, replayable, and produces evidence the next layer cites — so the final SAR is defensible all the way back to the source record.
- 1
Ingestion
Transaction monitoring, KYC, device, sanctions and adverse-media feeds
- 2
Context-gathering agent
Customer profile, transaction history, prior alerts, network neighbours
- 3
Typology-identification agent
Ranked typologies with citations to the typology library
- 4
Narrative-drafting agent
Jurisdiction-correct SAR structure, evidence-linked sentences
- 5
Quality-scoring agent
Usability rubric, rewrite suggestions, defensibility check
- 6
Review & approval workflow
Investigator → MLRO → export, immutable audit trail
Compliance
Compliance is a first-order product feature.
VigilPath operates on data that is itself regulated. Every architectural decision — residency, tenancy, model contracting, audit — is made with that in mind.
- UK & EU data residency. London and Frankfurt regions of a Tier-1 hyperscaler. UK-incorporated controller, full UK GDPR posture.
- Strict tenant isolation. Row-level security enforced at the database; no investigator can read data outside their tenant.
- POCA s.333A tipping-off protection. Investigation material ringfenced from non-FC personnel; no customer-facing communications reference an investigation.
- LLM data handling. Frontier-model APIs contracted to prohibit training on customer data and retention beyond the processing call.
- Immutable audit trail. Every state transition, edit and export is logged with actor, timestamp and reason — exportable for supervisory review.
- SOC 2 Type II. Readiness by end of Year 2 per the published build plan; ISO 27001 alignment from day one.
- WCAG 2.1 AA. Keyboard-first interaction, screen-reader-friendly markup, sensible focus management throughout.
Set up a tenant in under two minutes.
Provision an isolated workspace and load a demo caseload to explore.