Skip to content
VICTA
/en/odvetvi/finance
industry

Finance.

Banks, insurers, fintech startups, leasing companies. We know what the Czech National Bank wants to see—and we write both marketing and code so a regulator won't flag it. We design AI use cases to be auditable from day one, not bolted on as a patch.
01 · problem

What clients bring us.

Banks, insurers, fintech startups and leasing companies face regulatory pressure from multiple directions at once: the Czech National Bank, ESA, AML 5/6, PSD2, DORA. The KYC process employs dozens of people manually reviewing documents and approving clients by hand. Fraud detection generates thousands of false positives a day that the compliance team has to investigate.

AI looks like a solution, but the regulator demands auditability, explainability and control over the data. Generic US cloud services are off the table. Marketing has to comply with legal requirements on investment advertising, which slows campaigns by weeks. Internal systems are old, but replacing them is a multi-year project.

02 · approach

How we do it.

We build AI in finance to be auditable from day one: every model has documentation, monitoring, an explainability layer and a fallback to a human decision. KYC automation through OCR and face matching, with manual review only for borderline cases—not full outsourcing of the decision. For AML and fraud detection we use explainable ensemble models (SHAP, LIME), so the compliance team understands why a model flagged a transaction.

We handle the marketing side just as carefully: copy goes through compliance review as part of the CMS workflow, not at the end. We build websites, investment calculators and portal solutions with the Czech National Bank, GDPR and PSD2 in mind. All on Czech or EU-jurisdiction infrastructure.

03 · how we proceed

4 steps from audit to growth.

  1. 01

    Regulatory and data audit

    3 weeks: Czech National Bank, AML and GDPR requirements, mapping existing systems, data sensitivity. We define red lines and audit-trail requirements.

  2. 02

    KYC and fraud pilot

    10–16 weeks: OCR + face matching for new clients, an ensemble model for fraud detection. A pilot on a subset of clients with a manual double-check.

  3. 03

    Monitoring and compliance

    Sentry-style monitoring of models (drift, accuracy, fairness), audit logs, integration with reporting tools. An annual model review before audit.

  4. 04

    Marketing and portals

    In parallel: website, calculators, client portal—all within a compliance workflow. Conversion measurement, A/B tests that respect regulatory constraints.

next step

A free call comes first.

30 minutes, no strings attached. We'll talk through what you're dealing with in your industry and recommend the next step—for larger projects, that can mean a paid audit of your whole stack.