AI for Financial Services: Knowledge assistants, ESG scoring and document workflows – on infrastructure your compliance team already trusts

The short answer

AI for financial services means three things in practice: knowledge assistants over policy and product documentation, ESG and sustainability scoring, and document triage for inbound forms. All three are in production for clients operating under German financial regulation. We do not do credit scoring AI, algorithmic trading or model-risk applications – those are explicitly high-risk under the EU AI Act and outside our scope.

What this means in practice

GDV (German Insurers Association) runs our AI Knowledge Assistant over tens of thousands of policy documents for 400+ member companies managing more than 500 million insurance contracts. Researchers find what they need in half the time, shadow AI use has dropped, and internal employee satisfaction is up. The system runs on Azure AI Search + GPT-4o via Microsoft AI Foundry – auditable, cited, EU-hosted.

Evangelische Bank runs a production sustainability scoring B2B tool that aligns customers' businesses with the Paris 1.5°C climate target via XDC equivalence metrics. Built on bank-compliant Kubernetes infrastructure with Vue + TypeScript frontend, Node + TypeScript backend, Postgres, and full audit trails – 6–9 months Scrum from prototype to production.

For inbound documents – claims forms, member enquiries, credit application pre-screening – the same pattern as our innatura email agent applies: classify, route, draft a reply for mandatory human review. No automated outbound decisions; every action is a draft awaiting sign-off.

Key components

Outcomes

What running compliant AI in production actually delivers.

Regulation to know before you start

DORA (Digital Operational Resilience Act) – in force since January 2025. ICT third-party risk, incident reporting and operational resilience testing now apply to AI systems. Any AI tool that processes financial data is a DORA-relevant ICT service. Our default EU-hosted stack gives you the data residency and audit trail DORA requires.

EU AI Act – credit scoring is high-risk (Annex III, point 5b). AI systems that assess creditworthiness, insurance risk or investment suitability require a conformity assessment, human oversight mechanism and audit log. We scope this in every discovery workshop and will tell you clearly if your use case falls here.

BaFin AI expectations. BaFin has published supervisory guidance covering model validation, explainability and model risk management. A knowledge assistant that cites its source document for every answer is inherently more explainable than free-text generation. We build citations in by default.

How it works

Why N3XTCODER

We bring a decade of impact-tech experience and over 160 AI projects since 2019. We run scoping sessions and build engagements that ship:

• GDV (German Insurers Association) – AI Knowledge Assistant over tens of thousands of policy documents for 400+ member companies. Research time halved, shadow AI use dropped, employee satisfaction increased. Azure AI Search + GPT-4o via Microsoft AI Foundry.
• Evangelische Bank – production sustainability scoring B2B tool aligning customers with the Paris 1.5°C target via XDC metrics, on bank-compliant Kubernetes infrastructure.
• A leading German association – AI Member Platform combining chat-based discovery with traditional category filters, on Microsoft AI Foundry + pgvector.
• innatura – AI email agent classifying enquiries and drafting replies with mandatory human review, in pilot on n8n and Azure OpenAI.
• Default stack: n8n in Berlin, Qdrant in the EU, Azure OpenAI via Microsoft EU Sovereignty, plus open-source EU alternatives like Mistral and Milvus on request.

Honest constraints

Credit scoring, insurance pricing and investment suitability AI are high-risk under the EU AI Act. These use cases require conformity assessment, audit logs and human oversight mechanisms beyond what we scope in a standard build engagement. We will tell you clearly if your use case falls here.

BaFin expects explainability. A system that generates an answer without citing its source does not satisfy regulatory expectations. Grounded RAG with citations is our default; we do not ship free-text generation in regulated contexts.

Banking secrecy (§ 203 StGB) and data sovereignty. Client data must not leave EU infrastructure. Our default stack keeps all processing within EU borders and can be configured for on-premises deployment where required.

Frequently asked questions

Other Services