Home

Financial crime intelligence  ·  Banks & governments

Your systems see transactions.We see the network behind them.

Civitas connects the systems you already run, resolves the same person or company across all of them, and builds one graph of accounts, contracts and payments — then scores the networks that matter and shows investigators exactly why.

Request a briefing

Everything else — platform, worked example, security, contacts — is in the menu, top right.

Entity graph · live view
Cluster 07 · risk 87 · high
Person Company Account Flagged network

02 — Platform

One engine.
Two very different jobs.

The intelligence core is shared. The integrations, rules, models and investigator workflows are built for the institution using them.

    Seven stages, in order

    01

    Connect

    Ingest from core banking, AML, fraud, KYC, procurement and case systems — without replacing any of them.

    02

    Resolve

    Match the same person, company or account across sources, spelling variants and stale records.

    03

    Graph

    Build one graph of people, companies, accounts, contracts and transactions, every relationship typed.

    04

    Detect

    Run rules, anomaly detection and graph-level models over behaviour, not single payments.

    05

    Prioritize

    Score entities and networks so the queue starts with what deserves an investigator's day.

    06

    Explain

    Show the factors, weights and underlying records behind every score. No unexplained verdicts.

    07

    Learn

    Feed investigator decisions back in, so confirmed cases sharpen the model and false positives quiet it.

    Where it sits

    Systems you already run
    Core bankingAML systemsFraud systems KYC / KYBGovernment databasesProcurement & paymentsCase management
    The Civitas layer

    It connects them, resolves the entities, and builds the intelligence graph.

    ConnectsReads every source through one ingestion layer.
    Resolves entitiesOne identity per person and company, across systems.
    Builds the graphTyped relationships between everything it ingests.
    Adds AI signalsAnomaly, classification and graph-level risk models.
    Correlates activityLinks behaviour across accounts, contracts and time.
    Gives contextEvidence, relationships and next steps per alert.
    Records outcomesEvery decision captured, attributed and auditable.
    Feeds backOutcomes retrain the models that raised the alert.
    What investigators get
    Prioritized alertsExplained risk scoresNetwork view Supporting transactionsRecommended next stepsFull audit trail

    03 — Worked example

    Six hops from a clean
    supplier to a closed loop.

    No single payment here is unusual. The shape they form is. Step through a procurement network the way the graph assembles it.

    One person and four companies linked by ownership, shared directors, tender bidding and payments, forming a closed loop. owns shared director bids against pays traces back to TENDER AWARDED CLOSED LOOP · RISK 87 · HIGH PERSONA COMPANYA COMPANYB COMPANYC COMPANYD
    Hop 01 / 06

    01 / 06

    04 — Inside a case

    A score is only useful
    if it can be argued with.

    Every score breaks down into factors. Every factor links to the records that produced it. An investigator can agree, disagree, or throw it out — and the system learns from that.

    Entity · supplier

    Company ABC Trading Ltd

    87
    / 100 · high
    LowElevatedHigh

      Each factor opens the underlying records — the tender documents, the registry filings, the individual payments. Nothing in the score is unsourced.

      Investigator workflow
      • Alert raisedNetwork risk crosses threshold — a pattern across five entities, not one transaction.
      • Case createdOpened automatically, with the entities, period and triggering activity attached.
      • Summary writtenA plain-language account of what was found and which relationships drove the score.
      • Graph and evidence attachedNetwork view, supporting transactions and every contributing factor in one place.
      • Next steps suggestedWhat to check first, based on which evidence would most change the picture.
      • Investigator decidesThe system recommends. A person decides — and that decision is the training signal.
      DismissMonitorInvestigateEscalate

      Every meaningful action — who opened it, what they saw, what they chose — is written to an audit log you can hand to a regulator.

      05 — Security & compliance

      Built for the strictest
      room in the building.

      Banks and government agencies don't get to relax their controls for a vendor. Neither do we.

      Encryption at rest & in transitIAM / RBACSecrets management Secure APIsNetwork segmentationImmutable audit trails Data classificationVulnerability managementIncident response Private or in-tenant deployment

      Boundaries we hold

      What Civitas will not do

      Being clear about the limits is part of being credible in this market.

      • No autonomous accusations. The system produces evidence and a recommendation; a human makes the finding.
      • No replacement for investigators. It prioritizes their attention — it does not substitute for their judgement.
      • No production citizen or customer data during evaluation. Pilots run on synthetic or de-identified records.
      • No unexplained scores. If a factor can't be traced to a record, it doesn't reach an investigator.
      • No boiling the ocean. We deploy against a named fraud typology first, then extend.