Start a pilot
Fraud & AML transaction monitoring

Every decision, explained.

Real-time, fully explainable fraud and AML decisions, from transaction to lodged SMR, in one platform. Every transaction gets a KScore in under 200 ms, and every score breaks down into named reasons that you, your auditor and AUSTRAC can read.

The problem

Scam liability is rising on one side and AUSTRAC scrutiny on the other. Detection vendors and case-management vendors each sell half the answer. Black-box ML makes it worse, because "the model scored it low" is not an answer a regulator accepts.

No ML in the decision path

Two transparent layers decide, and every decision decomposes into its parts.

Rules

Per-client policy packs: velocity, structuring, new-payee and round-amount patterns, KYC/KYB status, sanctions and PEP screening, jurisdiction and corridor limits. Versioned, and every change is backtested on your own history before it goes live.

Network graph

Fraud rings, mule chains, dispersal, layering and shared-identity links: the coordination that per-account anomaly tools miss.

ML and spectral graph analysis run in shadow only. They earn a place in decisions with measured results, the same standard a regulator holds you to.

Detection to lodgement, one system

One append-only audit trail. A regulator walkthrough is a login, not a document hunt.

AlertScored in real time
CaseFour-eyes review
CDDFiles, sanctions & PEP screening
SMRStatutory clocks
LodgedReceipt on file

Onboarding is covered too: business applicants are pre-vetted (company and directors screened, risk-rated) before you pay for full KYB, and the onboarding risk rating carries through to transaction monitoring.

Why it's different

01

Explainable by construction

Every decision names the rules and graph signals that drove it.

02

Network-native

Rings, mules and shared beneficiaries are explicit graph structure.

03

Detection to lodgement

Competitors stop at the alert. KScore carries it to the regulator.

04

Backtest before enable

Every rule change previews its impact on real history first.

05

Built for your jurisdiction

Home country, base currency, business calendar and reporting regime set per client, with FX normalisation built in.

06

Safe when things fail

An outage is never treated as evidence and never silent. Every affected score is stamped.

Proof points

  • Live in production, scoring real card and payments traffic, with a hard sub-200 ms path for card authorisation.
  • Multi-tenant in production, and able to score for a client whose banking platform runs in its own AWS account.
  • Compliance layer live: CDD files, beneficiary and customer screening, SMR register with deadline tracking.

Your data stays yours

A complete, isolated stack per client in AWS Sydney: its own database, graph, cache, keys and user pool. Clients share code, never data. One API in, decisions and webhooks out — hosted by us, or deployed into your own AWS account.

ShadowWe prove it on your traffic
ReviewYou supervise every hold
EnforceIt decides, at your pace

Start with a shadow pilot

You provide a historical extract and a live or batch feed. We provide day-one baselines from your own history, weekly calibration reviews, and at the end your would-have-been alert queue, SMR candidates and a side-by-side against your current tooling.

6–8 weeksNo decisioningNo customer impactFixed fee
Talk to us — hello@kscore.ai