H2O.ai Solution Briefs — Australian Financial Services & Public Sector

Brief 1 of 3 · Fraud Detection

Stop Fraud Before
It Costs You More

Australian banks and government agencies lose billions annually to fraud. Static rule engines and siloed tools cannot match the speed or sophistication of modern fraud vectors. H2O.ai's unified predictive and agentic AI platform stops fraud in real time — with the explainability regulators require.

70%

Reduction in scam losses at Commonwealth Bank of Australia using H2O.ai real-time AI

80%

Fewer false alarms — operational teams investigate real threats, not noise

<50ms

Real-time transaction scoring — fraud decisions made at the speed of payment

4–8wk

Typical time from PoC to production deployment with Grandmaster-led delivery

The Problem

Why Existing Fraud Controls Are Failing

Fraud in Australia has accelerated sharply. Scam losses to Australian consumers and businesses exceeded $3 billion in 2023 (ACCC Scamwatch). Authorised payment fraud, synthetic identity fraud, and account takeover are outpacing the detection capabilities of legacy rules-based engines — which are slow to update, generate excessive false positives, and provide no explainability for compliance teams or customers.

Rules engines lag reality

Fraud patterns mutate faster than rule libraries can be updated. Criminals exploit the gap between each update cycle.

False positives erode trust

Overfitted models block legitimate transactions, creating customer friction and support cost without reducing genuine fraud.

Explainability gaps

Black-box scores cannot satisfy ASIC/APRA audit requirements or support customer-facing dispute resolution under the ePayments Code.

H2O.ai Platform Capabilities

Predictive + Agentic AI — Working Together

Predictive AI

Real-Time Transaction Scoring

H2O Driverless AI builds GBM and deep learning ensemble models that score every transaction in <50ms. AutoML continuously retrains on new fraud patterns. Built-in SHAP explainability provides reason codes for every decision — audit-ready by design.

Agentic AI

Intelligent Case Investigation

H2O AI Super Agent orchestrates multi-step investigation workflows: pulling transaction history, cross-referencing behaviour patterns, summarising evidence, and drafting SAR narratives — reducing analyst time per case from hours to minutes.

GenAI

Customer Communication & Scam Detection

h2oGPTe analyses inbound communications and payment narratives in real time to detect social engineering and authorised push payment (APP) scam indicators — adding a language-based signal layer that numeric models alone miss.

Proven Outcomes

70%

Scam loss reduction — Commonwealth Bank of Australia

80%

Fewer false alarms on fraud alerts

95%

Proactive model drift detection before performance degrades

50%

Infrastructure cost savings vs. fragmented point solutions

H2O.ai | Fraud Detection Solution Brief

Australian Financial Services & Public Sector

How It Works

Fraud Intelligence Across the Payment Lifecycle

1

Ingest & Feature Engineering

H2O Feature Store centralises transaction, behavioural and device signals. AutoML builds hundreds of fraud-specific features automatically.

2

Real-Time Scoring

Ensemble models score transactions in <50ms. SHAP reason codes returned with every prediction for operator transparency.

3

Agent-Led Investigation

Flagged cases routed to AI agents that compile evidence dossiers, cross-reference history, and recommend action.

4

Continuous Learning

MLOps monitors for drift. Models retrain automatically when fraud patterns shift — no manual intervention required.

5

Audit & Governance

Full decision lineage, model version control, and automated compliance reporting — ready for APRA CPS 230 and ASIC oversight.

Why H2O.ai Wins This Use Case

Four Differentiators That Matter

Proven at Australian Scale

CBA — Australia's largest bank — runs H2O.ai in production for real-time fraud. Not a pilot. Not a proof of concept.

Explainability Built In

SHAP values on every prediction. Reason codes for operators and customers. Satisfies APRA, ASIC and ePayments Code obligations.

Sovereign Deployment

On-premise, private cloud, or air-gapped. Your transaction data never leaves your environment. SOC 2 Type II certified.

One Platform — No Stack Sprawl

Predictive AI, GenAI, agents, MLOps and governance in a single licensed platform. No per-transaction fees. Cost certainty.

Platform Components Used

What Runs for Fraud

H2O Driverless AI H2O AI Super Agent h2oGPTe H2O Feature Store H2O MLOps H2O Document AI

Target Personas

Chief Risk Officer / Group Executive Risk — accountability for fraud losses and regulatory outcomes
Head of Financial Crime — operational ownership of detection and investigation
CTO / Chief Data Officer — platform architecture and AI strategy
APRA-regulated ADIs — Big 4 banks, regional banks, credit unions
Federal & State Government — payments fraud in Centrelink, ATO, Services Australia

⚡

Regulatory Context — Australia

APRA CPS 230 Operational Risk · RBA Payment System Reforms · Scams Prevention Framework Act 2025

Australia's Scams Prevention Framework (2025) places direct liability on banks to prevent scam losses. APRA CPS 230 requires demonstrable operational risk controls. H2O.ai's explainable, auditable AI models provide the evidence trail regulators require — and the detection performance to reduce mandatory reimbursement liability under the SPF.

Brief 2 of 3 · AML & Counter-Terrorism Financing

AML/CTF Compliance
That Scales With Reform

Australia's AML/CTF Act amendments (Tranche 2) are the most significant financial crime reforms in two decades — extending AUSTRAC obligations to lawyers, accountants, real estate agents and other newly regulated sectors from 2026. H2O.ai helps both existing reporting entities and newly regulated businesses build AI-native compliance programs that detect, explain and report financial crime at scale.

2026

AUSTRAC Tranche 2 compliance deadline — designated non-financial businesses must be registered and compliant

90%

Effort reduction in compliance pipeline automation using H2O.ai MLOps and agent-driven workflows

100x

Document processing capacity — H2O Document AI ingests KYC, beneficial ownership, and transaction records at scale

24/7

Continuous monitoring — AI agents watch transaction patterns around the clock with no analyst fatigue

The Regulatory Imperative

AUSTRAC Tranche 2 — What Changes and When

The Anti-Money Laundering and Counter-Terrorism Financing Amendment Act 2024 extends Australia's AML/CTF regime to designated non-financial businesses and professions (DNFBPs) — including legal practitioners, accountants, real estate agents, precious metal dealers and trust and company service providers. Obligations include customer due diligence, transaction monitoring, suspicious matter reporting (SMR), and threshold transaction reporting. Compliance programs must be enrolled with AUSTRAC by mid-2026 with full obligations active from July 2026.

New entities starting from zero

DNFBPs face building AML programs without prior compliance infrastructure, expertise, or technology — in a condensed timeline under AUSTRAC scrutiny.

Existing programs under pressure

Banks and fintechs face higher AUSTRAC expectations on transaction monitoring quality, SMR timeliness, and network-level typology detection — while managing legacy system debt.

Explainability is not optional

AUSTRAC requires audit trails for monitoring decisions. Black-box models cannot satisfy regulatory expectations or support SMR narrative quality.

H2O.ai Platform Capabilities

AI Purpose-Built for AML/CTF Compliance

Predictive AI

Transaction Monitoring & Typology Detection

H2O Driverless AI builds ensemble models tuned to AML typologies — layering, structuring, smurfing, trade-based money laundering — with SHAP explainability that maps model decisions to specific regulatory risk indicators. Continuous retraining adapts to evolving criminal methods.

Agentic AI

Automated SAR/SMR Drafting & Case Management

H2O AI Super Agent compiles alert evidence, searches customer history, cross-references beneficial ownership and sanctions lists, then drafts Suspicious Matter Report narratives aligned to AUSTRAC reporting standards — reducing analyst write-up time by over 70%.

GenAI

Document Intelligence & Network Analysis

H2O Document AI extracts structured data from corporate structures, trust deeds, and beneficial ownership registers. h2oGPTe synthesises disparate data sources to surface hidden relationship networks and entity linkages that rule-based systems miss entirely.

Compliance Outcomes

90%

Reduction in manual pipeline effort via MLOps automation

100x

Document processing capacity for KYC and beneficial ownership

25%

Improvement in SMR/SAR quality scores through AI-assisted drafting

99.9%

Model version reliability in production with full audit trail

H2O.ai | AML & CTF Solution Brief

Australian Financial Services & Public Sector

🏛

AUSTRAC Reform — Tranche 2 Key Facts

AML/CTF Amendment Act 2024 · Designated Non-Financial Businesses & Professions · Compliance From July 2026

Tranche 2 extends AML/CTF obligations to ~100,000 additional Australian businesses. Regulated entities must implement: (1) AML/CTF programs with risk assessments; (2) customer due diligence and ongoing monitoring; (3) threshold transaction reporting (TTR) for $10,000+ cash; (4) suspicious matter reporting (SMR) within 24 hours of suspicion forming; and (5) record-keeping for 7 years. AUSTRAC has signalled active enforcement — penalties for non-compliance include civil penalties up to $18.5M per contravention and criminal liability for serious breaches. H2O.ai provides the technology backbone to meet all five obligations at scale.

How It Works

From Transaction Event to AUSTRAC Report — Automated

1

Data Ingestion

Connect transaction systems, KYC databases, SWIFT, beneficial ownership registries — via pre-built connectors. No rip-and-replace.

2

Typology Detection

Driverless AI scores every transaction against AML typology models. Network graph AI surfaces hidden entity relationships across customer portfolios.

3

Alert Triage

AI agents prioritise alerts by risk score. Low-risk alerts auto-dispositioned. High-risk alerts escalated with evidence pre-populated.

4

SMR Drafting

AI agents draft Suspicious Matter Report narratives, cross-checked against AUSTRAC guidance and reporting standards — ready for analyst review.

5

Audit & Reporting

Full decision trail stored. Automated TTR/SMR submissions. Model explainability reports produced on demand for AUSTRAC inspections.

Why H2O.ai For AML/CTF

Differentiators That Matter to Compliance Teams

Explainable By Design

Every model decision has SHAP-based reason codes. AUSTRAC examiners can inspect what drove a detection — not just that it fired.

Sovereign Deployment

Customer financial data stays in your environment. On-premise, VPC, or air-gapped. No third-party data exposure — critical for DNFBP legal privilege concerns.

Fast Time to Compliance

Pre-built AML typology model templates and Grandmaster-led delivery. New entities can reach compliance readiness in 8–12 weeks.

Proven Global Financial Institutions

Goldman Sachs, Wells Fargo, TD Bank, CBA, UOB, Banco do Brasil — H2O.ai runs in production across the world's most regulated financial institutions.

Platform Components Used

What Runs for AML/CTF

H2O Driverless AI H2O AI Super Agent H2O Document AI h2oGPTe H2O Feature Store H2O MLOps

Target Personas

Chief AML Officer / MLCO — accountability for AUSTRAC compliance program
Head of Financial Intelligence — SMR quality, typology coverage, detection performance
CTO / CDO — platform architecture, data sovereignty, integration
Legal, Accounting & Real Estate Firms — Tranche 2 newly regulated, building programs from scratch
AUSTRAC-supervised banks & fintechs — upgrading legacy transaction monitoring systems

Brief 3 of 3 · Perpetual KYC (pKYC)

Replace Periodic Reviews
With Continuous Intelligence

Traditional KYC operates on fixed review cycles — annual, triennial, or risk-tier based. In that interval, customer risk can change materially and go undetected. Perpetual KYC (pKYC) replaces point-in-time reviews with continuous AI-driven monitoring that triggers refresh workflows only when risk signals change — reducing cost while improving compliance quality and AUSTRAC audit readiness.

60%

Reduction in KYC refresh cost — targeted event-driven reviews replace blanket periodic cycles

100x

Document processing capacity — H2O Document AI extracts and validates KYC evidence at machine speed

40%

Improvement in customer risk accuracy — continuous signals vs. static periodic snapshot data

7yrs

AUSTRAC record-keeping obligation — full customer audit trail maintained automatically by H2O.ai platform

The Problem

Periodic KYC Is an Expensive Compliance Theatre

Australian banks, credit unions, and Tranche 2 entities face a common dilemma: periodic KYC reviews are costly, disruptive for customers, and still miss risk changes that occur between cycles. A customer whose beneficial ownership structure changes, who appears on a new adverse media report, or whose transaction behaviour shifts dramatically may not be reviewed for 12–36 months under current practices. AUSTRAC's enhanced CDD requirements — and the incoming Tranche 2 obligations — demand a higher standard of ongoing monitoring.

Cycle-based reviews miss real-time risk

Customer risk changes continuously — sanctions lists update daily, beneficial ownership structures change, adverse media emerges. Annual reviews are blind to the interval.

Review backlogs are a regulatory liability

Overdue periodic KYC reviews are one of AUSTRAC's most cited compliance failures. AI-driven prioritisation and automation eliminates the backlog permanently.

Manual processes can't scale for Tranche 2

DNFBPs cannot replicate bank-scale KYC teams. AI is not optional — it is the only viable path to compliant ongoing monitoring at the volumes required.

H2O.ai Platform Capabilities

The AI Engine Behind Perpetual KYC

Predictive AI

Continuous Risk Scoring & Change Detection

H2O Driverless AI builds customer risk models that run continuously — scoring each customer against behavioural baselines, sanctions updates, PEP list changes, and adverse media signals. Risk scores update in near real-time. Refresh workflows trigger only when scores change materially, not on calendar schedules.

Agentic AI

Autonomous KYC Refresh Workflows

H2O AI Super Agent orchestrates the full CDD refresh cycle: collecting updated documentation, running identity verification checks, resolving discrepancies, escalating edge cases to analysts, and updating customer risk profiles — with a complete audit trail, without human initiation.

GenAI + Document AI

Document Extraction & Verification

H2O Document AI and H2OVL Mississippi extract structured data from identity documents, company registers, trust deeds and beneficial ownership filings. h2oGPTe synthesises adverse media, sanctions databases and open-source intelligence (OSINT) into structured risk summaries — without manual analyst research.

pKYC Outcomes

60%

KYC operational cost reduction through event-driven refresh

40%

Risk accuracy improvement over periodic snapshot approach

100x

Document processing scale for CDD evidence extraction

Zero

Overdue KYC review backlog — continuous monitoring replaces periodic queue

H2O.ai | Perpetual KYC (pKYC) Solution Brief

Australian Financial Services & Public Sector

🔄

Regulatory Driver — AUSTRAC Ongoing CDD & Tranche 2

Ongoing Customer Due Diligence · Enhanced CDD for High-Risk Customers · FATF Recommendation 10

AUSTRAC's AML/CTF Rules require reporting entities to conduct ongoing CDD proportionate to the risk a customer presents. For high-risk customers — PEPs, high-transaction entities, complex structures — enhanced CDD requires more frequent review. The Tranche 2 reforms extend these obligations to DNFBPs. FATF's Recommendation 10 demands continuous monitoring of business relationships. H2O.ai's pKYC architecture directly operationalises these requirements — replacing manual periodic reviews with continuous, documented, AI-driven monitoring that satisfies AUSTRAC examination standards.

How It Works

Perpetual KYC — Event-Driven, AI-Orchestrated

1

Continuous Signal Ingestion

Sanctions lists, PEP registers, adverse media, ASIC registers, beneficial ownership — monitored continuously via automated feeds into H2O Feature Store.

2

Risk Score Update

Driverless AI recalculates each customer's risk score when new signals arrive. Score deltas above threshold trigger review workflow automatically.

3

AI-Driven Refresh

AI Agent orchestrates CDD refresh — collecting documents, verifying identity, resolving discrepancies, and updating risk profile with minimal analyst effort.

4

Analyst Review

Complex cases escalated to analysts with full evidence pre-assembled. Analysts review and approve — not research and compile. Decision time reduced 70%+.

5

Audit-Ready Record

Every risk score change, document collected, and decision made is logged automatically. 7-year AUSTRAC record-keeping met without additional effort.

Why H2O.ai For pKYC

Built for Regulated Environments

Continuous, Not Calendar-Driven

Risk scores update when signals change — not when a clock ticks. High-risk customers get more attention; stable low-risk customers cost less to maintain.

Explainable Risk Scores

Every customer risk rating has a SHAP-based explanation — what drove the score up or down. Analysts and examiners can interrogate any decision.

Sovereign Architecture

Customer PII and financial data stays in your environment. Critical for DNFBP legal professional privilege and banking data sovereignty requirements.

One Platform — No Vendor Sprawl

Feature Store, risk models, document AI, agent orchestration, and audit trail — all in H2O.ai. Annual license, no per-transaction fees. Cost certainty for compliance budgets.

Platform Components Used

What Runs for pKYC

H2O Driverless AI H2O AI Super Agent H2O Document AI h2oGPTe H2O Feature Store H2O MLOps H2OVL Mississippi

Target Personas

Chief AML Officer / MLCO — accountable for CDD program quality and AUSTRAC outcomes
Head of KYC Operations — owns the review cycle, backlog, and operational cost
CTO / CDO — responsible for AI platform architecture and data governance
Legal and Accounting Firm Managing Partners — Tranche 2 compliance owners in newly regulated entities
Real Estate Group Risk & Compliance — building pKYC for high-value property transactions