Use Case Insurance · Schwabing Insurance Cluster · BaFin + EIOPA

Solvency II + IFRS 17 + EU AI Act in one Decision Layer - Schwabing insurance cluster Munich Re + Allianz

Munich Re + Allianz in the Schwabing insurance cluster. 740+ QRT data points per quarter. BaFin MaGo VAG + VAIT for insurers. Internal Model governance with audit trail before BaFin. UK PRA SS1/23 parallel.

Workshop at Grindelberg All 7 use cases ←

Chapter 1 - Three Supervisory Regimes, One Decision-Layer

Solvency II + IFRS 17 + EU AI Act simultaneously - but per step deterministically assigned.

Munich insurers in the Schwabing Königinstraße cluster (Munich Re Königinstraße 107, Allianz SE Königinstraße 28 - 800 m apart) carry three parallel supervisory worlds: Solvency II (BaFin mandatory reporting, EIOPA validation rules per quarter, 740+ data points), IFRS 17 (since 2023 for listed and group-consolidated insurers - BBA + CSM + Risk Adjustment), EU AI Act (Annex III Point 5(c) for life/health insurance, plus GDPR Art. 22 for every automated underwriting decision).

Plus regulatory: BaFin MaGo VAG (Circular 2/2017, Minimum Requirements for the Business Organisation of Insurance Undertakings) and VAIT (Insurance Supervisory Requirements for IT, Circular 11/2017 revised) - mandatory model governance + Use-Test for Internal Models used in Solvency II SCR calculation. BayLDA AI Checklist v0.9 from 24.01.2024 for Munich supervision. DSK Orientation Guide AI and Data Protection 06.05.2024. 2022 Allianz BaFin order (Structured Alpha Fund, Internal Controls) as precedent for public supervisory requirement. UK parallel: UK insurers face PRA SS1/23 model risk management + Solvency UK + IFRS 17 - parallel architecture demands.

Decision-Layer split typically for claims triage + underwriting + QRT reporting: 40% RULES (EIOPA validation rules per data point, IFRS 17 BBA logic, cession limits, treaty scope), 35% AI AUTONOMOUS (claims-file classification, Cat modelling range, language versions on international claims), 25% HUMAN (underwriter/actuary discretionary decisions, large losses above value thresholds, model-validation audit sign-off).

Audit trail per decision: model version (e.g. cat_model_v3.7), EIOPA rule version (e.g. eiopa_qrt_v2026_q2), IFRS 17 BBA assumption set, human override + reasoning + timestamp. At a BaFin audit or EIOPA stress test, the path crosses the table, not the result. 1-click export for BaFin format + EIOPA format + internal PwC/EY/Deloitte audit view.

Chapter 2 - Decision Record for a Munich Re Cat Modelling Decision

How a natural-catastrophe portfolio assessment runs in the Decision-Layer.

Anonymised decision record for a Cat modelling decision at Munich Re reinsurance acceptance. Tropical cyclone exposure in the Caribbean. Treaty renewal preparation 2026. AI-supported Cat modelling with BaFin VAIT + EIOPA Use-Test + GDPR + EU AI Act logging.

CAT-2026-05-17-CARIB-TC-RNW

Cat modelling treaty renewal · Caribbean tropical cyclone · 47 treaty cases with layer structure · received 17.05.2026 · Solvency II SCR relevant

Result Pricing range generated · Underwriter sign-off mandatory
  1. 01 REGEL

    Treaty eligibility check

    47 treaty renewal cases in the Caribbean region. Per treaty: layer structure (Excess of Loss, Quota Share, Aggregate Excess) with individual cession limit verification. Treaty scope (Caribbean, Tropical Cyclone Cat 3+) mapped against internal underwriting guidelines. Rule treaty_elig_v3.4.

    ✓ 47 treaties layer-structured
  2. 02 REGEL

    EIOPA QRT capacity check

    Current Solvency II SCR utilisation 73%. Remaining Cat risk budget: EUR 480M aggregate. Rule eiopa_qrt_v2026_q2.

    ✓ Budget available
  3. 03 KI

    Cat modelling (model <code>cat_model_v3.7</code>)

    RMS + AIR + Verisk models calculated in parallel. Tropical Cyclone Cat 3+ Caribbean region. Three separate sizes: Treaty median EAL EUR 142M (median single-treaty contribution across 47 layer-structured treaties). Portfolio aggregate EAL EUR 380M (not 47 × 142 = 6.7bn, because cession cross-effects + reinstatement stress + multi-treaty correlations are deducted). 1-in-200 PML EUR 2.4bn (Munich Re Caribbean exposure scale). Model ensemble confidence 0.91. Climate-change adjustment +12% (2026 study set).

    Confidence 0.91 · threshold 0.85

    ✓ Treaty median 142M · Portfolio 380M · PML 2.4bn
  4. 04 REGEL

    EIOPA Use-Test (Internal Model)

    Model must actually be used for pricing (Use-Test per Solvency II Art. 120 + EIOPA Guidelines on Internal Models). Cat model output feeds directly into pricing engine. Use-Test logging active. Rule solvency2_use_test_v1.0.

    ✓ Use-Test confirmed
  5. 05 KI

    Pricing range generation

    Cat model output + treaties + cession limits → pricing range per cedant. Model pricing-engine-v4.2. Min/max spread calculated. Climate-change premium considered.

    Confidence 0.88 · threshold 0.85

    ✓ Pricing layer per treaty individual
  6. 06 MENSCH

    Underwriter sign-off on large contracts

    Mandatory stop on treaties with layer limit > EUR 200M (24 of 47). Senior Underwriter Mr K. (Cat-Treaty specialist, 17 years Munich Re) receives the decision record with Cat model output, pricing proposal, cession utilisation. Per large contract: manual assessment + sign-off. Confidence override possible + documented.

    ✓ 24 large contracts signed
  7. 07 REGEL

    GDPR Art. 22 auto-decision check

    For 23 treaties below EUR 200M layer limit: AI pricing range accepted automatically. For AI pricing decisions with significant cedant impact: GDPR Art. 22 right to challenge documentation generated. Rule art22_auto_v1.4.

    ✓ Right-to-challenge documented
  8. 08 REGEL

    IFRS 17 + Solvency II double-booking

    Treaty renewal output booked simultaneously into IFRS 17 BBA (Best Estimate + Risk Adjustment + CSM) and Solvency II SCR position. Versioned rule sets in parallel: ifrs17_bba_v3.2 + sii_scr_v4.1. Consistency technically enforced.

    ✓ Both standards current
  9. 09 REGEL

    Audit trail persist (BaFin + EIOPA + IFRS 17 + EU AI Act)

    Complete decision record persisted. 1-click export for BaFin VAIT audit format (Use-Test view), EIOPA QRT format, IFRS 17 audit view (PwC/EY/Deloitte), EU AI Act Art. 12 logging format. Rule audit_v1.4.

    ✓ Audit trail persisted

Chapter 3 - Workshop at Munich Urban Colab or at Munich Re/Allianz

Engineering from Hamburg, workshop at the Schwabing insurance cluster.

Engineering head office Hallerstraße 8 Hamburg, Munich workshop on-site. Discovery workshop can be at Munich Urban Colab (Freddie-Mercury-Straße 5, Kreativquartier) or directly at Munich Re/Allianz in Schwabing. Separate rooms for CRO session, actuarial workshop, compliance/DPO briefing, works-council session. Workshop under EUR 10,000. Engineering phase remote with bi-weekly on-site days.

Schwabing insurance cluster advantage: Königinstraße 28 (Allianz) and 107 (Munich Re) are 800 m apart - parallel discovery workshops for both groups are possible. Common compliance standard because of subsidiary relationships (Allianz Re is a Munich Re competitor, both are BaFin-supervised). GDPR consensus for the Munich insurance market through BayLDA AI Checklist 01/2024. UK parallel: workshop in English with remote bridge to UK head office (Lloyd's, Aviva, Prudential).

Integration with insurance IT: Decision-Layer integrates with insurance platforms: SAP Insurance Module, Allianz Internal Suite, Munich Re Cat modelling tools (own IT plus RMS, AIR, Verisk), Solvency II reporting suites (BearingPoint Abacus, Moody's RiskCalc). REST/SOAP adapters depending on group system. Source code of the adapters is handed over with the repository to Munich Re/Allianz - no vendor lock-in.

Actuarial team pattern: Munich insurance actuarial functions are among the most experienced worldwide. Decision-Layer is introduced as an 'actuarial tool with audit trail', not as 'AI replaces the actuary'. Works council expert right per § 80 (3) BetrVG (German Works Constitution Act) is respected - IG Metall or ver.di experts are included in the workshop. Group works agreement (Konzernbetriebsvereinbarung) template with measurable escalation thresholds for treaty renewal seasons (January renewal round + April Cat treaty). UK equivalent: ICE Regulations + works-council consultation + ACAS Code on AI in workplace decisions.

Related use cases

Frequently asked questions

Which Munich insurers does this spoke address?
Munich Re (Königinstraße 107) and Allianz SE (Königinstraße 28) - both in the Schwabing insurance cluster, about 800 m apart. Plus Allianz subsidiaries Allianz Global Investors (AGI) and Allianz Investment Management. Solvency II ratio 2024: Munich Re 300%, Allianz 277%. BaFin direct supervision. Plus EIOPA stress-test obligation (most recent 2024, next 2027). Plus IFRS 17 since 2023 for listed insurers. UK insurers face PRA SS1/23 model risk management + Solvency UK + IFRS 17 - parallel architecture demands.
How is the 740+ QRT data-point problem addressed with the Decision-Layer?
Solvency II QRT (Quantitative Reporting Templates) have over 740 data points per quarter, typically including 38 manual overrides from claims files. At Munich Re/Allianz with 90,000 employees globally that is a 4-day sprint with 22 actuarial staff doing overtime at quarter-end. The Decision-Layer extracts structured data from surveys, claims reports, expert opinions - and maps them against EIOPA validation rules. Manual overrides remain a human duty (escalation), but every position gets an audit trail with rule version. EIOPA validation-rule updates per quarter are stored as versioned rule sets. UK parallel: PRA SS1/23 + Solvency UK demand the same auditability per QRT/MCR position.
What do BaFin MaGo VAG + VAIT say about AI models at insurers?
BaFin supervises insurers via MaGo VAG (Circular 2/2017) and VAIT (Circular 11/2017, revised), alongside the EIOPA stress-test obligation and Internal Model approval. Together these tighten model governance for Solvency II Internal Models and their AI risk components: a model must be validated, backtested and actually used (the Use-Test), not merely documented. The 2022 Allianz Structured Alpha Fund case showed BaFin can publicly require internal controls. The Decision-Layer contributes an audit trail per model decision - model version, input hash, confidence score and Use-Test logging - with DPO and CRO sign-off paths at validation audits. The UK equivalent is PRA SS1/23 model risk management, with FCA Consumer Duty for retail components.
How is the IFRS 17 vs. Solvency II double-booking problem solved?
IFRS 17 (mandatory since 2023 for listed insurers) requires a different valuation than Solvency II - Building Block Approach for Best Estimate, Risk Adjustment, Contractual Service Margin. Actuarial teams at Munich Re + Allianz rebuild Excel models quarterly for both standards. The Decision-Layer stores both valuation logics as versioned rule sets in parallel - one input (portfolio, claims, discount rates), two outputs (IFRS 17 position, Solvency II SCR). Consistency is technically enforced, audit trail documents per position which assumptions were used. PwC/EY/Deloitte audit-ready in one click export. UK insurers face IFRS 17 in parallel - the same dual-output architecture applies.
Does the spoke also address underwriting AI for Munich Re reinsurance?
Yes. Munich Re reinsurance acceptance (Treaty and Facultative) uses AI for natural-catastrophe (Cat) modelling, pricing decision support and cession optimisation. For life and health insurance, EU AI Act Annex III Point 5(c) classifies risk assessment and pricing as high-risk. P&C and reinsurance are not explicitly high-risk, but GDPR Art. 22 and BaFin VAIT model governance still apply. In the Decision-Layer, rules cover cession-limit and treaty-scope checks, AI produces the Cat-modelling and pricing ranges, and a human is mandatory on large losses above value thresholds and on underwriting discretion. The UK equivalent is PRA SS1/23 for UK-domiciled reinsurance entities and the Lloyd's market.

Schedule workshop at Grindelberg

3-day discovery: Day 1 process analysis, Day 2 Decision-Layer mapping, Day 3 use-case prioritisation. Concrete deliverable.

Schedule meeting

Discovery workshop below EUR 10,000. Pilot fixed price discussed after the workshop.

← Back to overview: AI Agents Hamburg (7 use cases)