Banker-X emblem: a bank, an AI chip, a shield and a stylised X

Banker-X

Agentic AI-Powered Trust Layer for Banks and Government

Safe banking. Inclusive finance.

One evidence layer for banks and government agencies: who a person or business really is, where they are, who they are connected to, and whether the proof is genuine. Agentic AI proposes; a person always approves.

Request a pilot conversation Where we are today

Institutions cannot tell who is real, where they are, how accounts connect, or whether the evidence is genuine

Hidden networks

Fraud rings and shell entities split value across many accounts. Each looks ordinary alone.

A number is not evidence

A valid BVN or NIN proves a record exists, not that the person is present or where they live.

Evidence can be faked

Photos, scans and video can be edited or AI-generated, and the eye cannot tell.

Real people are turned away

Women and rural citizens without formal ID or address cannot complete onboarding or reach services.

Five working engines behind one layer

Identity Link

BVN, NIN and phone checks with consent. A BVN record is released only by the person's own one-time code. Match flags only; nothing is stored.

AdresX

Signed, verifiable address evidence. Built for basic phones and field agents.

AML-X

Sanctions and PEP screening, alerts, cases and report drafting, plus verification of agencies and companies.

Deepfake-X

Screens photos, scans, video and audio for manipulation or AI generation. Scores are indicative; an analyst reviews them.

360GIL

Shows where women are, and are not, included in finance.

Agentic AI: agents call the engines as tools, and a person approves every action.

Three working agents, a fourth planned

Onboarding, investigator and inclusion agents each have one role and a fixed set of tools, and work end to end on a synthetic bank. A trust agent is planned.

Human approval

The model proposes, the engines return evidence, and a person decides. An agent never approves its own work.

Audit trail and limits

Every step is recorded, with data minimisation and spend and rate limits built in.

One platform, two editions

For banks and fintechs

  • Consented identity and address evidence at onboarding
  • Screening of ID photos, selfies and documents for manipulation
  • Sanctions, PEP and adverse-media screening
  • Alert triage and report drafting; mule-ring analysis (in development)
  • Inclusion analytics for lenders and regulators

For government agencies

  • Identity and address evidence for public services
  • Verification of agencies, companies and counterparties
  • Authenticity screening of photo, scan and video evidence
  • Entity risk screening for compliance and trade
  • Case management with an audit trail; agent analysis approved by an officer

Both editions share one core: the evidence layer, the agent runtime and the governance controls. Government and bank data are designed to be kept in separate tenants.

A government example: customs risk profiling

Live

High-risk declarations

Verify and screen the importer, consignee and agent, and link them by shared addresses, phones and owners.

Live, indicative

Altered documents

Screen supporting photos, scans and video for manipulation or AI generation.

Graph view in development

Non-compliance patterns

Officer case files with evidence and an audit trail, with link analysis across entities.

Needs agency data

Misclassification and valuation

Models trained on an agency's own declaration data under an MoU. Not built yet.

We show what we do today and what needs the agency's data, so that you can judge us on facts.

Where we are today

Live
  • AML-X: screening, alerts, cases
  • Identity checks: BVN with consent, NIN, phone
  • Deepfake-X: image, video and audio screening
  • 360GIL, built for the AIR × CBN sprint
  • NDPC-registered data controller
In testing
  • A Tier-1 Nigerian bank is testing AML-X
  • Security, defence and financial-crime agencies hold test accounts
  • NIPOST postcode link on the test environment
  • CBN Regulatory Sandbox application submitted
  • Agentic AI: three working agents on synthetic data
Building now
  • The Banker-X case console
  • A mule-ring graph view
  • Deepfake-X checks inside the agent flow

We separate what is live from what we are building, so that you can judge us on facts. All demonstrations use synthetic data.

The team

Timothy Avele

Founder and CTO. Builds and operates the AGXII platform suite for Nigerian defence, law-enforcement and government agencies, including AML-X, AdresX and 360GIL. Cybersecurity, OSINT and digital forensics consultant to NCTC/ONSA.

Rapheal Crossdale

AI Safety and Governance. Leads how the agents are constrained, approved and audited, and how data from banks and government agencies is kept apart.