CapabilityAI for AML

    AML case preparation, with the analyst in charge

    Short answer

    An AML harness: an AI agent gathers the evidence, traces the funds and drafts the case file inside a written mandate. A compliance analyst reviews it and decides. Every step is logged with the case.

    AML harnessEvery case
    Mandatewritten list of allowed sources and actions
    Decisionalways the analyst, never the agent
    Evidenceevery statement linked to its source
    Action logevery step stored with the case
    Datastays in your environment
    How an AML case movesThe analyst decides
    1. 01Alert inmonitoring or screening · manual referral
    2. 02Evidence gatheredcustomer and transactions · on-chain traces
    3. 03Draft case filetimeline · linked sources
    4. 04Analyst decidesclose, escalate or report · never the agent
    5. 05Loggedevery step · ready for audit

    Demo

    From a stolen-funds report to a draft case

    Three steps on a sample case: the client describes what happened and gives the transaction, the agent traces the funds hop by hop, and the analyst gets a draft report to check and sign.

    Investigation demo · three sample cases

    Pick a case

    What the client reported

    The victim's seed phrase was compromised. USDT was swept from the wallet to an address that already had other victim reports, then moved on to an exchange.

    Read-only sample. Addresses masked.
    Case intakeRead-only
    NetworkTRON
    AssetUSDT
    Start tx6227f6f2…2e01
    Stolen14,342.78 USDT

    How one case runs

    1. Alert in

      An alert arrives from your monitoring system, screening tool or a manual referral, with the case data you already hold.

    2. Evidence gathering

      The agent pulls customer data, transaction history, screening hits and on-chain traces from the sources it is allowed to read.

    3. Draft case file

      It writes a draft with a timeline, the evidence behind each statement and the questions still open. Every source is linked.

    4. Analyst decides

      A compliance analyst reviews the draft, edits it and makes the decision: close, escalate or report. The agent never files on its own.

    5. Logged

      Every agent step, source, edit and decision is stored with the case, ready for audit or regulator review.

    Division of labour

    The analyst keeps every decision

    The agent may

    • Read the case data and sources named in its mandate
    • Trace funds on chain and summarise counterparties
    • Draft the case narrative with linked evidence
    • Flag gaps and suggest the next check

    The agent may not

    • Close, escalate or report a case
    • Contact a customer or a counterparty
    • Read systems outside its mandate
    • Change a decision once the analyst signs

    Venture

    Built and run by us

    Built by Vareger founders

    An AI agent that gathers evidence for anti-money-laundering reviews.

    • Gathers evidence only from the sources its written mandate names, and records each source.
    • Runs the checks the compliance team already uses and drafts the case.
    • Decides nothing alone: the analyst signs every decision.
    • Every step, input, source and result is stored with the case.
    Built byVareger founders
    ProvesAgents for regulated work, with a person signing every decision.

    Common questions

    Q01Does the agent decide on cases?

    No. It prepares the case file. A named analyst reviews it and makes every decision, and the log shows who decided what.

    Q02Which data sources can it use?

    Only the ones written into its mandate during scoping, for example your case management system, transaction data, screening results and blockchain analytics you already license.

    Q03How does a pilot work?

    We start with a written scope on a sample of your past alerts. The agent drafts cases, your analysts compare them with their own work, and we measure time per case and rework before anything touches live work.

    Q04Where does the data go?

    The agent runs in your environment or in a dedicated one we set up for you. Case data is not used to train models.

    Next step

    Test it on your past alerts

    A scoping call with Oleg. The pilot runs on a sample of your closed cases before anything touches live work.

    Oleg Pun

    CEO, Vareger

    Oleg Pun, CEO, Vareger