AI for financial-crime compliance
An Australian entity intelligence graph with AI investigation.
Averis Graph connects public-source intelligence into a reusable graph for KYC screening, then uses AI agents to help resolve potential screening matches with evidence.
Intelligence should compound. Investigation should not.
Entity intelligence · product demonstration
Connected intelligence
7 links
Claim
Held public office: 2016–2019
Source
Parliamentary record (illustrative)
Evidence
“David Lee, Member for Western Metropolitan Region. Term commenced March 2016, concluded March 2019.”
Australia
Depth before breadth. The initial focus is Australian intelligence, with international coverage developed progressively.
Initial focus
Coverage expands source by source and jurisdiction by jurisdiction. It is not yet comprehensive across Australian or international risk intelligence.
Long-term vision
The strategic asset is the intelligence graph — not an individual screening workflow.
Today's typical workflow
Screen → Alert → Research → Search → Investigate → Decide
The Averis direction
Enrich → Resolve → Evidence → Decide
Where we are today: an MVP that builds source-linked entity intelligence and supports investigator match decisions. Monitoring and infrastructure-scale services are planned, not yet available.
The problem
Incomplete records create alerts that look defensible — and investigations that start from almost nothing.
Screen
A financial institution screens a customer.
Match
Similar names generate potential matches.
Research
An investigator searches sources and compares identities.
Resolve
Conflicting or missing information must be resolved.
Document
The decision and its evidence must be documented.
Repeat
Then again for the next customer and each later review.
A screening alert is a lead to investigate — not proof of identity or wrongdoing.
Where the cost comes from
Vendor and platform costs
Screening data and software
Repeated investigations
People’s time: search, compare, document
High total compliance cost
Averis Graph aims to reduce duplicated research and unnecessary investigative effort. Lower total cost is the thesis we are testing with design partners — not a proven result.
The Averis principle
Enrich the intelligence record before asking the customer record to match it.
Build a richer, source-linked picture of the person or organisation first. Then use that context to assess a potential customer match.
The intelligence graph
A name tells you who might be a match. Averis builds the context around that name — connecting identities, roles, relationships, events and evidence over time, with every material claim traceable to its source.
Provenance path
Relationships, events and provenance — structured over time. Explore the intelligence model →
Evidence and provenance
Averis Graph preserves not only what is known, but why. Investigators can trace a conclusion back through the claim and evidence to its source.
Provenance trail
Entity
David Lee
The person the claim describes
Claim
Held public office: 2016–2019
Support status: Supported · Confidence: High
Evidence
“Member for Western Metropolitan Region, 2016–2019.”
The specific extract that supports the claim
Source
Parliamentary record
Where the information came from · date observed
Contradiction retained · date of birth
Claim A
1962
Supported · source-linked
Claim B
1964
Contradictory · surfaced, not overwritten
AI agents
Research agents
Find relevant information in permitted sources and capture facts with their source references and observation context.
Enrichment agents
Help connect information to the right entities, add context and surface conflicting claims or missing attributes.
Human accountability
Agents gather and organise evidence; validation controls graph changes, and the accountable investigator makes the final disposition.
Investigation agents
Compare a customer with candidate entities, gather evidence for specific questions and organise supporting, contradictory and missing information.
Shared memory
Every agent reads from and adds to the same graph, so research done once can be reused by later investigations instead of being reconstructed from scratch.
Commercial wedge
Averis Graph is designed to sit alongside existing screening and case-management platforms as an intelligence and investigation layer. The initial proposition is not to replace an institution's compliance stack.
Economic thesis · to validate
Lower effort per resolved candidate — and potentially lower total compliance cost.
This is the thesis we are testing with design partners, not a proven result.
Founder
20 years in financial services and consulting · financial-crime compliance
Michael Phillips brings 20 years of financial-services and consulting experience across Australia and the UK, with deep expertise in financial-crime compliance. His career includes financial-crime leadership and MLRO roles at Bolt, HUB24 and Airwallex; major financial-crime screening, transaction-monitoring and payments-screening initiatives at Westpac; and financial-crime advisory work at Deloitte.
He founded Averis Graph to tackle a problem he understands firsthand: the high cost of screening data and the substantial manual effort required to investigate potential matches and document defensible decisions.
michael@averisgraph.comBolt2025–
Senior Manager FCC & MLRO
HUB242022–25
Head of FCC & MLRO
Airwallex2021–22
Deputy MLRO
Westpac2018–21
Senior Manager, FCC
Deloitte2012–18
Financial Crime Manager
Accenture UK2005–10
FS Consultant
MLRO = Money Laundering Reporting Officer · FCC = Financial-crime compliance
Design partners
We are seeking financial-services design partners to test evidence quality, investigation effort and cost per resolved candidate on representative screening cases. Institutional enquiries about enrichment, evidence and integration are welcome.
hello@averisgraph.commichael@averisgraph.com
Intelligence should compound. Investigation should not.