For VARA · ADGM FSRA · DFSA · CBUAE licensees

Second-line AML, done in hours instead of days.

finaml.ai is an AI case agent for UAE-regulated VASPs and fintechs. It triages alerts, builds the case file, drafts the narrative and prepares goAML STRs, with every step logged for VARA, ADGM and CBUAE audits. Your MLRO approves. Nothing is filed without a human.

Human approval on every action Full source-linked audit log UAE data residency
app.finaml.ai/cases/CASE-2026-00417
finaml.ai case workspace: grouped alerts, funds-flow graph, counterparties and the AI narrative draft with citations

The problem

Screening tools raise the alert. Everything after that is still done by hand.

Hours per case

Gathering KYC, transactions and screening hits from five systems, then writing the narrative and preparing the filing, is manual work on every case.

Most alerts are noise

Analysts spend the bulk of their time closing false positives, while the cases that matter wait in the same queue against the same SLA.

STR quality is under scrutiny

UAE supervisors increasingly test the quality, sourcing and timeliness of suspicious transaction reports, not just whether they were filed.

What finaml.ai does

From alert to filed report, one case file.

01

Triage

Groups related alerts and pulls KYC, transactions and screening hits into one case. Suggests a priority and SLA.

Agent · analyst can override
02

Investigate

On-chain and fiat flow analysis across wallets and IBANs. Matches activity against typologies mapped to UAE regulator guidance.

Analyst + agent
03 AI output

Draft

Writes the case narrative and a disposition recommendation. Every sentence carries a citation; every recommendation carries a confidence level.

Agent · analyst edits, edits stored as a diff
04

File & respond

Builds the goAML STR with schema validation and prepares regulator RFI response packs. Filing requires MLRO approval.

MLRO · four-eyes on every filing

Product

Built around the case, not the alert.

Narrative draft with numbered citations, confidence level and recommended disposition
Narrative with citationsEach sentence links to the alert, transaction or document behind it. The analyst edits inline and picks a disposition.
goAML STR builder mapping case transactions to goAML fields with validation
goAML STR builderReport, subjects, transactions and narrative are pre-filled from the case and validated against the UAE FIU schema before submission.
Append-only audit log of human and model actions
Audit log and decision recordsAppend-only, hash-chained log of every human and model action, with a read-only auditor role.
Product walkthrough

Follows one case through all 25 screens, from sign-in and data sources to the regulator's request and the audit record.

Open the walkthrough

Built for UAE supervision

Aligned to the regulators you report to.

Typologies, filing formats and record-keeping follow the rulebooks your MLRO is examined against.

VARAADGM FSRADFSACBUAEUAE FIU goAML

Typologies mapped to UAE guidance

Each typology cites the VARA, FSRA, DFSA or CBUAE rule or guidance note it derives from. Custom typologies can be added for the entity's own risk assessment.

goAML XML schema validation

STRs are validated against the current UAE FIU goAML schema before an MLRO sees the submit button. Gaps are flagged per field.

Arabic / English documents

Trade licences, IDs and bank letters are read in either language and cited in the original.

Audit trail by design

Every recommendation can be replayed and explained to a supervisor.

Each AI action writes a log entry with the exact prompt version, the sources it read, the raw model output, the analyst's edits as a diff, and who approved it. Auditors and regulators get read-only access to the same record.

Decision log · CL-2041 · entry 14sha256:9b1f…e42a
Action
AI draft  Case narrative generated
Prompt version
narrative/v2.3.1 · model finaml-v2.3
Sources used
[1] Ledger · 4 inbound txns[2] Chainalysis KYT alert 88213[3] Sumsub KYC file 11 Mar 2025
Model output
“…61% of which traces within 2 hops to a mixer designated by OFAC [2]…” (412 words, stored verbatim)
Human edit
- Pattern is consistent with layering.+ Pattern matches VARA typology T-07 (layering via mixer exposure).
Approver
S. Haddad (MLRO) · second approver after R. Ahmed (Analyst)
Timestamp
2026-09-09T15:12:44+04:00

Integrations

Works with the tools already generating your alerts.

CSV and API import are available from the first day of a pilot. Vendor connectors are enabled per entity, and every connected source becomes a citable reference in the narrative.

Sumsub
ComplyAdvantage
Chainalysis
Elliptic
TRM Labs
Fireblocks
Onfido
Core ledger via API
CSV upload

Security

Customer data stays in the UAE and out of the model.

Separation of duties is enforced by the system, not just documented. What the agent may do unattended, and what always needs a person, is a written policy that is versioned like code.

Data residencyUAE region for production deployments
Model trainingNever on customer data
Single sign-onSSO / SAML 2.0
Access controlRole-based: Analyst, MLRO, Auditor
Audit logAppend-only, hash-chained
Model & prompt versionsPinned and logged per decision
Pilot environmentsIsolated · synthetic or anonymised data
SOC 2 Type IIRoadmap

Pricing

Priced by case volume, not by alert.

Pilot pricing on request · billed monthly · excl. VAT

Starter

For a first compliance hire and an MLRO.

On request
  • Up to 300 cases / month
  • 2 seats
  • Triage, investigate, draft
  • Full audit log
Request a pilot

Enterprise

For groups, banks and multi-licence operators.

Contact us
  • Unlimited cases and seats
  • Private deployment
  • Custom typologies
  • Dedicated compliance engineer
Contact sales

Pilots for VARA-, ADGM- and CBUAE-licensed entities are open now.

One case file, one reporting entity, four weeks. We set up the workspace, connect your alert sources and run alongside your team before anything is filed.