Glosár pre AT, sponzorov a future re-use
Tri vrstvy slovníka — Allianz-specific termíny, technický slovník, AI-specific glossary — plus mapovanie klientových materiálov a externé regulatory referencie. Niektoré Allianz termíny sú stále TBD a čakajú na confirmation z business teamu.
Domain glossary pre Allianz Slovensko
Allianz-specific termíny extrahované z prezentácie a Replit demo materialov. 5 z nich má status TBD a sú flag-nuté cez OQ-5 pre confirmation z Allianz business teamu.
| Termín | Význam | Status |
|---|---|---|
| Unius | Core poistný systém (policy admin), pravdepodobne posAm-built | TBD · OQ-5 |
| C360 | Customer 360 — unified data view per customer | OK |
| NPS | Net Promoter Score (customer satisfaction metric) | OK |
| DWH | Data Warehouse | OK |
| VPP | Všeobecné poistné podmienky (general policy conditions) | OK |
| APK | Analýza potrieb klienta (client needs analysis) | OK |
| AMC | TBD | TBD · OQ-5 |
| ZoR | TBD (Zápisnica o Rozhovore? Zmluva o Realizácii?) | TBD · OQ-5 |
| ID | TBD (Investičný Dotazník? Identifikačné Dáta?) | TBD · OQ-5 |
| Analýza PK | TBD (Analýza Poistných Kontraktov? Analýza Potrieb Klienta?) | TBD · OQ-5 |
| Karta klienta | Customer card (UI view) | OK |
| Karta agenta | Agent card (Phase 3 — agent's KPI/progression view) | OK |
| Kampaňový manažment | Campaign management | OK |
| posAm | Slovak system integrator, builds custom enterprise systems for FSI | OK |
| AT (Allianz Technology) | Wholly-owned IT subsidiary of Allianz Group, Munich-based | OK |
| OneTrust | Privacy + AI risk assessment platform used by Allianz | OK |
| AI Community | Allianz regional AI governance / model risk team | OK |
| ADP | Allianz Data Platform (group data platform initiative) | OK |
| Nedokončenky | Incomplete contracts / drafts | OK |
| Provízie | Commissions | OK |
Tech glossary — MCP, RAG, DORA a ďalšie skratky
Technické termíny použité naprieč briefom — od MCP architektúry cez Azure landing zone konvencie po regulatory frameworks.
| Termín | Význam |
|---|---|
| MCP | Model Context Protocol — Anthropic-developed open protocol pre AI model ↔ tools/data integration |
| MCP Host | Process that runs the LLM and orchestrates MCP clients |
| MCP Client | Per-server connection within MCP Host |
| MCP Server | Provides tools, resources, prompts to LLM via MCP protocol |
| RAG | Retrieval-Augmented Generation — augmentácia LLM výstupu o relevantný retrieved context |
| RSC | React Server Components |
| CDC | Change Data Capture |
| DPIA | Data Protection Impact Assessment (GDPR požiadavka) |
| DORA | Digital Operational Resilience Act (EU 2022/2554, FSI ICT) |
| ZDR | Zero Data Retention (Anthropic enterprise feature) |
| Hub-spoke | Network topology with central hub VNet + spoke VNets |
| Landing zone | Azure-managed, governed deployment environment |
| CAF | Microsoft Cloud Adoption Framework |
| WAF | Well-Architected Framework |
| SAST/DAST | Static / Dynamic Application Security Testing |
| SBOM | Software Bill of Materials |
| PrivLink | Azure Private Link |
| APIM | Azure API Management |
AI termíny — prompt caching, model tiering, agentic patterns
AI-specific koncepty, ktoré sú kľúčové pre cost engineering, agentic architecture a security narrative briefu.
| Termín | Význam |
|---|---|
| Prompt caching | Anthropic feature: 90% input token discount na repeated context (system prompts, tool definitions) |
| Model tiering | Strategicky využívanie cheaper modelov (Haiku) pre routing/classification, expensive (Sonnet/Opus) pre main work |
| Tool use | LLM ability to call external functions/APIs |
| Agentic patterns | Multi-step LLM workflows with planning, reflection, tool use |
| Embeddings | Numeric vector representation of text for semantic search |
| Vector store | Database optimized for vector similarity search |
| Hallucination | LLM generating plausible-sounding but factually wrong output |
| Jailbreak | Adversarial prompt trying to bypass LLM safety controls |
| Context window | Maximum tokens LLM can process in single request |
| Token | Sub-word unit; ~4 chars per English token, ~3-4 per Slovak |
Klientove materiály a kde v briefe sa používajú
Mapovanie 9 slidov z NextGen prezentácie + 12 screenshotov z Replit demo NextGenAppAgent → konkrétne sekcie deliverable-u.
NextGen Prezentácia — 9 slidov
Cesta: materials/NextGen_Prezentacia/
- Slide 1: MCP definition + workflow + architektúra → sekcia 3.3 + 5
- Slide 2: Current vs future state, dátové silá → C360 → sekcia 1.1, 5.4
- Slide 3: Schéma NextGen Agent — 3-phase visual → Big Picture + sekcia 5
- Slide 4: 1. fáza overview — C360 + Karta + Kampane → sekcia 5.1
- Slide 5: 1. fáza detail — Karta klienta + kampaňový manažment → sekcia 5.1
- Slide 6: Podporné moduly — Power BI, Outlook, AI asistent, Planner → sekcia 5.1
- Slide 7: 2. fáza overview — Documents + APK → sekcia 5.2
- Slide 8: 2. fáza detail — Documents + Silvia Fohl Belková (APK) → sekcia 5.2
- Slide 9-10: 3. fáza — AI Sales asistent + Karta agenta + X-sell → sekcia 5.3
NextGenAppAgent — 12 screenshotov
Cesta: materials/NextGenAppAgent/
Sponzorov early prototype z Replit-u, 12 UI screenshotov v range 08.19.24 – 08.20.27.
Primary use: sekcia 5.1 OQ-6 reconciliation — Replit demo malo features (Allie chat, kouč, vzdelávanie, rast, príležitosti, úlohy), ktoré z prezentácie patria do Phase 3.
Inspiration: UX patterns, navigation flows, agent-twin interaction model.
Detailed mapping každého screenshotu → feature → phase je TODO počas writing fázy.
Regulatory frameworks, vendor docs, industry research
Externé zdroje, ktoré tento brief cituje — od EU regulations cez Anthropic / Microsoft enterprise documentation po AI security frameworks.
| Source | Použitie v briefe |
|---|---|
| EU Regulation 2022/2554 (DORA) | Sekcia 6.5 regulatory matrix — ICT risk pre FSI od Jan 2025 |
| EU Regulation 2024/1689 (EU AI Act) | Sekcia 6.5 + AI risk classification — high-risk AI systems pre customer-facing CRM |
| EU Regulation 2016/679 (GDPR) | Sekcia 6.4 + 6.5 — DPIA, lawful basis, čl. 9 special category |
| EU Directive 2022/2555 (NIS2) | Sekcia 6.5 — cybersecurity controls per Annex II |
| Solvency II Directive 2009/138/EC + EIOPA guidelines | Sekcia 6.5 — insurance prudential, operational risk + ICT risk components |
| Anthropic Enterprise documentation (ZDR, prompt caching, MCP) | Sekcia 3.3 + 6 + 7.3 — AI architecture + cost optimization + compliance |
| Anthropic Claude Code documentation | Sekcia 8 — AI-augmented dev methodology |
| Microsoft Azure Cloud Adoption Framework (CAF) for FSI | Sekcia 3 + 6 — landing zone alignment, governance baseline |
| Microsoft Well-Architected Framework | Sekcia 3 — coverage 5 pillars (cost, ops, performance, reliability, security) |
| Allianz Group Brand Manual (verejné PDF) | Sekcia 10 — visual & brand specifications, color palette, typography |
| METR (Model Evaluation & Threat Research) studies | Sekcia 8.2 — calibrated 3-5x dev acceleration claim |
| GitHub Copilot productivity research | Sekcia 8.2 — empirický baseline pre AI-augmented dev productivity |
| OWASP Top 10 for LLM Applications | Sekcia 6.5 + 8.1 — AI-specific security scanning, Tier 1 dev controls |
| MITRE ATLAS (adversarial AI threat framework) | Sekcia 6.5 — threat modelling pre adversarial AI scenarios |
Glosár + materials mapping + externé referencie sú navrhnuté pre re-use. Brief je enterprise-grade artifact sám o sebe — aj keby projekt nikdy nedosiahol delivery, tieto sekcie tvoria kompletnú strategickú a kontextuálnu bázu pre future Allianz / FSI / agentic AI projekty.