Solvency II Article 41: AI Tools for System of Governance and Actuarial Function Compliance

Article 41: The System of Governance Framework

Article 41 of the Solvency II Directive (Directive 2009/138/EC) requires all EU insurance and reinsurance undertakings to maintain an effective system of governance providing for sound and prudent management of the business. The system of governance must include at minimum four key functions: the risk management function (Article 44), the compliance function (Article 46), the internal audit function (Article 47), and the actuarial function (Article 48). These are not organisational departments — they are formal governance functions with defined responsibilities, independence requirements, and reporting obligations to the Administrative, Management or Supervisory Body (AMSB).

The fitness and propriety requirements of Article 42 apply to all persons running key functions. Every individual responsible for a key function — including the Head of Actuarial Function (HoAF) — must notify the supervisory authority in advance. The National Competent Authority (NCA) assesses fitness (relevant knowledge of actuarial and financial mathematics, commensurate with the nature, scale, and complexity of the business) and propriety (honesty and financial soundness). In practice, NCAs expect the HoAF to be a Fellow of a recognised actuarial association — FIA, FCAS, FFA, or an equivalent IAA member body. For Lloyd's syndicates operating under UK Solvency II-equivalent rules, the PRA and FCA apply equivalent requirements under the Senior Managers and Certification Regime (SMCR), with the HoAF captured as a Senior Management Function (SMF).

Outsourcing of key functions is permitted under Article 49 but subject to significant constraints. The undertaking remains fully responsible for the outsourced function and must maintain appropriate oversight. Material outsourcing arrangements must be notified to the supervisory authority. The undertaking cannot outsource in a way that materially impairs governance quality or the supervisory authority's ability to monitor compliance. For the actuarial function specifically, EIOPA has clarified that day-to-day actuarial calculations can be outsourced to consultants, but the HoAF role should not be outsourced in a way that removes substantive actuarial oversight from within the undertaking — a consultant engagement that is genuinely supervised and challenged by an internal HoAF is acceptable; a purely nominal internal HoAF who rubber-stamps external work is not.

Article 48: The Actuarial Function

Article 48 defines the specific mandate of the actuarial function. It is not a description of the actuarial team as a whole — it is a specific governance function with the following enumerated responsibilities under Article 48(1) of the Directive:

• Coordinate the calculation of technical provisions. The actuarial function is responsible for coordinating the end-to-end process, ensuring methods and assumptions are appropriate, and providing a signed opinion on the reliability and adequacy of the calculation. It does not necessarily perform every calculation personally.
• Ensure appropriateness of methodologies, underlying models, and assumptions. This requires the HoAF to challenge the method selection for each homogeneous risk group (HRG): is chain-ladder appropriate given data volume and stability? Is a frequency-severity approach more appropriate for long-tail lines? Does the lapse assumption for the life block reflect current experience?
• Assess the sufficiency and quality of data used in technical provision calculations. The data quality mandate covers completeness, accuracy, and appropriateness. Data limitations must be documented and their quantitative effect on the best estimate uncertainty must be assessed.
• Compare best estimates against experience. Back-testing of prior year technical provisions against actual claims emergence is a core actuarial function responsibility. Systematic divergences — actual development consistently above or below the prior best estimate — must be explained and incorporated into the current year methodology.
• Inform the AMSB of the reliability and adequacy of the calculation of technical provisions. This formal opinion to the AMSB is delivered via the Actuarial Function Report (AFR) at least annually.
• Express an opinion on the overall underwriting policy. The HoAF reviews whether the portfolio being written is consistent with the assumptions underlying the technical provisions — including the implications of portfolio mix shifts, entry into new lines, or changes in underwriting appetite.
• Express an opinion on the adequacy of reinsurance arrangements. The actuarial function assesses whether the reinsurance programme is appropriate for the risk profile of the undertaking, covering both risk transfer adequacy and counterparty credit risk on reinsurance recoverables.
• Contribute to the effective implementation of the risk management system. This is particularly relevant for ORSA (Article 45) — the actuarial function contributes risk quantification, stress scenario modelling, and the assessment of standard formula assumptions versus the undertaking's actual risk profile.

EIOPA Guideline 42 specifies that the HoAF must possess relevant actuarial experience — typically at least three years of professional practice in the relevant line of business. An individual qualified as an FIA or FCAS but with no life actuarial experience would not typically satisfy fitness requirements for the HoAF role at a life insurer, and vice versa. NCAs apply proportionality — at smaller undertakings, a broader background is accepted; at large composite insurers, the NCA may expect depth in both life and non-life actuarial practice.

The Actuarial Function Report

The Actuarial Function Report is the primary formal deliverable of the actuarial function under Solvency II. It is produced at least annually and presented to the AMSB. The AFR is distinct from the Regular Supervisory Report (RSR) submitted to the NCA and the Solvency and Financial Condition Report (SFCR) published publicly. The AFR is an internal governance document — though the NCA may request it as part of a supervisory review or thematic exercise.

EIOPA Guideline 52 (EIOPA-BoS-14/253) specifies the minimum content of the AFR. It must cover all tasks undertaken by the actuarial function, clearly identify deficiencies, and provide recommendations for remediation. In practice, the AFR contains the following sections:

• Technical provisions methodology by line of business / HRG. Description of the reserving method selected for each HRG (chain-ladder, B-F, Cape Cod, Clark growth curve, frequency-severity, individual case basis), rationale for each selection, changes from prior year, and limitations of the selected method. For P&C, this typically covers each EIOPA Schedule of Business line separately.
• Assumption documentation and validation. Claims development assumptions, expense loading, lapse and persistency (life), mortality and morbidity tables and adjustments (life), reinsurance credit risk. Each assumption must state the derivation basis — company experience study, industry benchmark, professional judgment — and show comparison to prior year.
• Data quality assessment. Completeness, accuracy, and appropriateness of data used in technical provision calculations. Data limitations and their quantitative or qualitative effect on best estimate uncertainty must be documented.
• Best estimate uncertainty assessment. The HoAF must describe the range of uncertainty around the best estimate. For P&C, this often uses a triangles-based uncertainty range; for life, scenario testing or stochastic modelling. The AFR must not present a single point estimate as if it is precise.
• Back-testing and comparison with prior year. Actual development versus prior year best estimate by HRG. Material divergences (positive or negative) must be explained and the implications for methodology or assumption selection documented.
• Underwriting policy opinion. The HoAF reviews whether portfolio changes (new lines written, exits from markets, changes in cedant quality) remain consistent with technical provision assumptions. This requires the HoAF to engage with underwriting leadership before issuing the opinion.
• Reinsurance adequacy opinion. Is the reinsurance programme sufficient to protect the undertaking against the scenarios underlying the SCR and ORSA stress testing? Does it provide genuine risk transfer? This assessment includes counterparty credit quality on reinsurance recoverables.

• "Draft the technical provisions methodology section of an Actuarial Function Report for a European P&C insurer under Solvency II. Lines of business per EIOPA delegated regulation Annex I: Motor Vehicle Liability (LoB 1), Other Motor (LoB 2), Fire and Other Damage to Property (LoB 7), General Liability (LoB 9). Methods applied: (1) Motor Vehicle Liability: incurred chain-ladder for accident years 2019–2022 (sufficient development credibility); Bornhuetter-Ferguson for AYs 2023–2025 (MTPL long-tail; chain-ladder unstable at early maturities); tail factor 1.015 based on MTPL industry development benchmarks; (2) Other Motor: paid chain-ladder; credible history to 84 months; no B-F required; tail factor 1.004; (3) Property (LoB 7): separate treatment of attritional losses (below €500K threshold) using paid chain-ladder, and large losses (above €500K threshold) using frequency-severity model; catastrophe loading applied per EIOPA non-life standard formula nat-cat factors for the relevant NatCat perils applicable to the portfolio; (4) General Liability: incurred B-F for all accident years given long-tail development; a priori loss ratios based on 5-year actuarial pricing indications; tail factor 1.045 based on European GL industry development benchmarks. For each LoB document: method selection rationale, changes from prior year methodology with reasons, key assumptions, and material limitations. Format in EIOPA-compliant AFR style, suitable for AMSB review and NCA inspection."
• "Draft the data quality assessment section of the Actuarial Function Report for the same European P&C insurer. Data sources and quality findings: (1) Claims data: sourced from CLAIMS_DB as at 31 December 2025; IT reconciliation to policy administration system confirms total case reserves match general ledger to within 0.1% by LoB; (2) Data issue — Motor LoB: 147 claims records have missing bodily injury severity coding (1.2% of open BI claims by count, 0.8% of open BI case reserves by value); these claims have been assigned the average severity for their accident year and development period as an approximation pending data remediation by IT; the actuarial function estimates the impact of this approximation as immaterial — less than €200K on a LoB 1 best estimate of €84M; (3) Premiums: gross written premium by LoB reconciled to the general ledger; cession data confirmed from reinsurance bordereau signed by the reinsurance accounting team; (4) Reinsurance recoverables: IBNR recoverables estimated by applying treaty structure to gross IBNR by LoB; no specific excess-of-loss claims currently in dispute with reinsurers; (5) Large claims: the General Liability LoB contains 3 open claims exceeding €2M each that are reserved on an individual case basis outside the triangles process; combined case reserves €9.1M; these claims have been reviewed individually by the HoAF and are excluded from the triangles to avoid distortion. Draft the EIOPA-compliant data quality section identifying each issue, its materiality, and the remediation recommendation."

AI Governance Under Article 41: The 2024 EIOPA Opinion

EIOPA's 2024 Opinion on Artificial Intelligence Governance (adopted October 2024) represents a significant extension of Article 41 system-of-governance requirements to AI and algorithmic models used in insurance undertakings. The Opinion is addressed to National Competent Authorities and is expected to be applied through supervisory dialogue and NCA-level guidance across the EU. The UK PRA has issued equivalent expectations through its AI governance framework update, extending existing Senior Management Functions accountability to AI model risk management.

For actuarial functions, the 2024 EIOPA Opinion has three material implications:

• Documentation of AI tools used in technical provision calculations. If an insurer uses ML models, AI-augmented development factors, or algorithmic tools in calculating best estimates, the actuarial function must document these in the AFR: model purpose, data inputs, validation methodology, limitations, and oversight process. This applies whether the model is developed in-house, embedded in a consultant tool (Milliman Mind, AXIS), or procured from a third-party data platform.
• Validation of AI model outputs. AI models used in SCR-relevant calculations — including those that generate best estimates feeding into the SCR — are subject to model validation requirements. For standard formula users, this falls under general model governance requirements of Article 41. The actuarial function must challenge and validate AI outputs with the same rigour applied to traditional actuarial models, including out-of-time back-testing and sensitivity analysis on key parameters.
• Accountability to the AMSB. The HoAF must be able to explain AI model outputs to the AMSB in terms that a non-specialist board member can assess for reasonableness. Where AI tools have been used in the provision calculation chain, the AFR must describe these tools and the HoAF's validation of their outputs. The HoAF's opinion on technical provision adequacy is not delegated to the AI tool — the HoAF remains fully accountable.

From a practical standpoint, using Claude to draft AFR narrative sections, ORSA narratives, or AMSB briefing notes falls within the same category as using a word processor or a prior-year template — it is writing assistance, not actuarial model output. The 2024 EIOPA Opinion is primarily concerned with AI tools embedded in the calculation of technical provisions themselves (ML-based claims development models, neural network severity models, algorithmic pricing models used within the TP calculation chain). Actuarial functions using such tools should document the governance framework, backtesting results, and model limitations in the AFR. Functions using Claude solely for documentation drafting need only ensure the HoAF reviews and certifies the content.

• "Draft the AI governance section of the Actuarial Function Report for a P&C insurer that uses two machine learning models in actuarial processes: (1) a gradient boosting model for motor claims severity prediction, used to supplement the triangles-based best estimate for the most recent 3 accident years where statistical development is immature (the model adds a €2.1M upward adjustment to the LoB 1 best estimate of €84M, representing 2.5%); (2) a natural language processing model for claims triage that classifies bodily injury claims by severity band (minor, moderate, serious, catastrophic), used to improve case reserve setting accuracy. Both models were developed by the in-house data science team and are used in production. Document: (a) Purpose and materiality of each model in the technical provision calculation chain — is each model material to the best estimate per EIOPA proportionality guidance; (b) Validation framework: the severity model was validated using an out-of-time test set (2022–2023 claims held out; model trained on 2018–2021). Report out-of-time RMSE and compare to a chain-ladder baseline; (c) Oversight process: actuarial function reviews model outputs monthly and documents material differences versus the triangles-based reserve exceeding €500K; (d) Identified limitation: the gradient boosting severity model was trained pre-2023 and may not adequately capture post-2023 social inflation trends — actual development has been 8% above model prediction in 2024–2025; remediation plan: retrain model incorporating 2023–2025 data by Q2 2026; (e) EIOPA 2024 Opinion compliance: confirm the documentation framework meets the AI Governance Opinion requirements. Format as an AFR subsection suitable for NCA inspection."

Own Risk and Solvency Assessment: Article 45

Article 45 of the Solvency II Directive requires every insurance undertaking to conduct an Own Risk and Solvency Assessment at least annually and immediately following any significant change in risk profile. The ORSA is a forward-looking internal assessment of the undertaking's own risk profile and the capital needed to support it — it is not a regulatory calculation but a management and governance tool that the AMSB must own, understand, and act upon. The ORSA report is submitted to the NCA as part of the supervisory process.

The ORSA has three core components under Article 45: (1) an overall solvency needs assessment — estimating the capital the insurer actually requires to support its risk profile, which may differ from the regulatory SCR where the standard formula does not fully reflect the undertaking's risks; (2) continuous compliance with capital requirements — demonstrating that the insurer has maintained the SCR and MCR throughout the year, and quantifying how quickly solvency could deteriorate under stress scenarios; (3) deviation from standard formula assumptions — where the undertaking's risk profile differs materially from standard formula assumptions, this must be assessed and the impact on capital requirements quantified.

The ORSA is as much a narrative document as a quantitative one. Supervisors and boards read it to assess whether management genuinely understands and actively manages its risk profile — not merely whether the numbers pass the SCR test. The qualitative sections — description of the risk appetite framework, forward-looking solvency narrative, stress scenario discussion, and strategic implications of the ORSA findings — are where Claude's drafting capability adds the most value for actuaries and risk officers.

• "Draft the forward-looking solvency projection section of an ORSA report for a mid-size European P&C insurer under Article 45. Solvency position at 31 December 2025: Own Funds €480M (Tier 1 €420M, Tier 2 subordinated debt €60M), SCR €310M, MCR €100M, Solvency Ratio 154.8%. Business plan 2026–2028: GWP growth +8% per year; target combined ratio 94%, 92%, 91% for 2026–2028; investment return 3.8% on matched fixed income portfolio (no equity exposure, no duration mismatch); reinsurance programme unchanged. Projected Solvency Ratios (management central case): 2026 148%, 2027 151%, 2028 158%. Stress scenarios modelled: (1) Major European windstorm event (1-in-100 year return period): net retained loss €120M in 2026, post-stress Solvency Ratio 118% — remains above 100% SCR floor but triggers board escalation threshold of 130%; (2) Reserve deterioration: +15% strengthening on prior year net reserves (€42M adverse development), post-stress ratio 132%; (3) Combined stress (windstorm + reserve deterioration simultaneously): post-stress ratio 94%, breaching the 100% SCR requirement and triggering regulatory intervention threshold. Draft: (a) the business plan solvency projection narrative explaining the methodology and key assumptions; (b) stress test results section including management actions available for each scenario (the windstorm combined stress triggers subordinated debt issuance or capital call from parent); (c) assessment of whether the current solvency position is adequate for the risk profile, noting the limited capital headroom above the 130% board appetite; (d) emerging risk paragraph addressing EIOPA AI governance requirements and the insurer's use of ML models in claims pricing."
• "Draft the SCR sensitivity analysis commentary section for the ORSA report. Standard formula SCR components at 31 December 2025: Non-Life Underwriting Risk €148M, Market Risk €62M, Counterparty Default Risk (reinsurance) €28M, Operational Risk €34M, Diversification benefit -€72M, Basic SCR €200M, SCR (after loss absorbing capacity of technical provisions and deferred taxes) €172M. Own Funds €310M. Solvency Ratio 180%. Sensitivity analysis results (impact on SCR and Solvency Ratio): (1) Premium volume growth +15% vs. plan +8%: non-life underwriting risk increases approximately €22M, Solvency Ratio 168%; (2) Reinsurance counterparty downgrade — primary XL carrier (€80M limit, current rating A-) downgraded to BBB: counterparty risk charge increases €9M, Solvency Ratio 174%; (3) Reserve stress — +10% strengthening on GL and Motor net reserves: reserve risk charge increases €14M, own funds decrease €14M, Solvency Ratio 162%; (4) Interest rate shock +200bps parallel shift: market risk charge reduces from €62M to €44M (portfolio duration 4.2 years, rate rise reduces bond values but liability discount benefit approximately offsets for this duration), net Solvency Ratio change approximately -2 points given the own funds impact on bond valuation. For each sensitivity: explain the directional effect on SCR components, the impact on solvency ratio, whether it breaches board escalation threshold of 130% or regulatory minimum of 100%, and whether management actions are pre-approved for the scenario."
• "Draft the reinsurance adequacy opinion section of the Actuarial Function Report. Reinsurance programme in force for the 2025 underwriting year: (1) Property per risk excess-of-loss: €3M xs €2M, with one reinstatement at 100%; lead reinsurer Swiss Re rated AA- by S&P; (2) Property catastrophe excess-of-loss: four-layer programme totalling €150M xs €50M; lead reinsurers Swiss Re (AA-) and Hannover Re (A+); provides protection to approximately the 1-in-200 year return period; (3) Motor bodily injury excess-of-loss: €5M xs €1M per occurrence; protects against large bodily injury claims; in the last 5 years 3 motor BI claims have exceeded €1M (largest €2.4M); (4) General Liability excess-of-loss: €10M xs €500K per occurrence; one reinstatement at 100%. Actuarial assessment for the AFR reinsurance opinion: (a) Catastrophe protection: internal RMS model (1-in-100 year gross PML €185M, net of cat XL €34M; 1-in-200 year gross PML €240M, net €54M — marginally above the top layer limit; HoAF recommendation: consider extending cat protection or increasing retention by 20%); (b) Motor XL adequacy: at the €1M retention, the programme has paid out on 3 of the last 5 years; frequency suggests the retention is near the market, but consider whether severity trends (social inflation in BI awards) justify lowering the retention to €750K at next renewal; (c) GL XL adequacy: €500K retention is low relative to the attritional frequency in the GL portfolio (approximately 40 claims per year exceed €100K); the HoAF recommends raising the retention to €750K at renewal to reduce cession costs while retaining meaningful large-loss protection; (d) Counterparty credit: all reinsurers rated A or better by S&P or A.M. Best; total reinsurance recoverable €28M; no material counterparty credit risk identified. Draft the formal reinsurance adequacy opinion in AFR style."
• "Draft an AMSB board briefing note on solvency position and capital management for Q3 2025. Format: 2–3 pages; non-technical language suitable for board members who are not actuaries. Key data: Year-end 2024 Solvency Ratio: 162%. Estimated Q3 2025 Solvency Ratio: 147%. Movement drivers: (1) Q2 2025 European windstorm losses: net retained loss €38M against cat budget of €25M — excess €13M reduces solvency ratio approximately 8 points; (2) GL reserve strengthening: €9M adverse development on two large bodily injury claims in litigation — reduces solvency ratio approximately 4 points; (3) Business growth: GWP +14% YTD versus plan +8% — SCR growth from volume absorbs approximately 5 points; partially offset by: (4) Investment return: portfolio returned 4.1% YTD (above 3.5% plan) — adds approximately 2 points. Net movement: -15 points from 162% to 147%. Actions: (a) Cat budget exhausted — next significant loss event comes entirely from capital; (b) GL remediation underway; (c) €30M Tier 2 subordinated debt issuance in process — requires AMSB approval; estimated impact +12 points (153% post-issuance). Board appetite minimum: 130%. Briefing note should: (1) explain the solvency ratio movement in terms a non-actuarial board member can assess; (2) confirm whether the current 147% ratio is adequate against board appetite; (3) set out the subordinated debt proposal and its effect; (4) provide the forward-looking solvency trajectory into Q4 2025 and the 2026 ORSA baseline."

Practical Article 41 Compliance Workflow

For the Head of Actuarial Function at an EU insurance undertaking, the Article 41 and Article 48 compliance cycle maps to the following annual governance calendar:

• Q1 (January–March): Produce the year-end Actuarial Function Report covering prior year technical provisions, data quality, reinsurance adequacy opinion, and underwriting policy opinion. Present to AMSB. Back-test prior year best estimates against actual development. Contribute actuarial input to the year-end SFCR and RSR disclosures.
• Q2 (April–June): Mid-year technical provision update if risk profile has changed materially (e.g., following large loss events, significant portfolio changes, or acquisition). Assess reinsurance renewal outcomes against the prior AFR reinsurance opinion. Begin ORSA data gathering — engage underwriting and finance for the business plan inputs that drive the forward-looking solvency projection.
• Q3 (July–September): ORSA stress testing and scenario analysis. Draft ORSA report for AMSB review and sign-off. Quarterly solvency monitoring report to AMSB. Update AI model validation if ML models are used in technical provision calculations — confirm that model performance has not drifted materially since last validation.
• Q4 (October–December): Year-end technical provision calculation begins — data extraction, triangle updates, method review. Any methodology changes from prior year require HoAF sign-off before calculation and AFR documentation. Reinsurance programme renewal — HoAF provides actuarial input to the reinsurance adequacy assessment for the coming underwriting year.

Claude with ClaudeFinanceLab's Insurance & Actuarial templates is most effective in the narrative-intensive phases: AFR section drafting, ORSA forward-looking narrative, AMSB briefing notes, and the reinsurance adequacy opinion. For Lloyd's syndicates operating under equivalent UK rules, the Lloyd's Syndicate Actuarial guide covers Syndicate Business Forecast commentary and Lloyd's Minimum Standards compliance documentation. The Life Insurance Reserving guide covers VM-20 principle-based reserving for US entities — which shares methodology parallels with Solvency II best estimate reserving for life and health lines. For a broader overview of AI tools available to actuaries, see Best AI Tools for Actuaries in 2026. For the core IBNR methodology that underpins the AFR's technical provision section, see the IBNR Calculation guide.

Frequently Asked Questions

What are the four key functions required by Solvency II Article 41?

Article 41 of Directive 2009/138/EC requires every EU insurance and reinsurance undertaking to maintain four key governance functions: (1) risk management function (Article 44), (2) compliance function (Article 46), (3) internal audit function (Article 47), and (4) actuarial function (Article 48). Each must be effectively independent with defined responsibilities and direct reporting to the AMSB. All persons running key functions are subject to the fitness and propriety requirements of Article 42 and must be notified in advance to the National Competent Authority. In the UK, equivalent obligations apply under the Senior Managers and Certification Regime (SMCR) with the Head of Actuarial Function captured as a designated Senior Management Function.

What must the Actuarial Function Report contain under EIOPA Guideline 52?

Under EIOPA Guideline 52 (EIOPA-BoS-14/253), the AFR must document all tasks undertaken and their results, identify deficiencies, and provide remediation recommendations. Minimum required content: technical provisions methodology and assumptions by homogeneous risk group, data quality assessment, back-testing of prior year best estimates against actual development, uncertainty assessment, underwriting policy opinion, reinsurance adequacy opinion, and the actuarial function's contribution to the risk management system and ORSA. The AFR is produced at least annually and presented to the AMSB. It is an internal governance document — it is not the SFCR (public) or the RSR (NCA supervisory report), though the NCA may request it.

Does EIOPA require insurers to document AI tools used in technical provisions?

Yes. EIOPA's 2024 Opinion on AI Governance extended the Article 41 system-of-governance requirements to AI and algorithmic models used in technical provisions, pricing, and claims. Actuarial functions using ML models in best estimate calculations must document these tools in the AFR — covering model purpose, validation framework, data governance, identified limitations, and oversight process. Using Claude to draft AFR narrative sections is writing assistance and falls outside the AI model governance requirements, which target AI embedded in the actuarial calculation chain. The HoAF remains fully accountable for technical provision adequacy regardless of AI tool involvement.

When is an interim ORSA required under Article 45?

Article 45 requires the ORSA to be conducted at least annually and immediately following any significant change in the risk profile. EIOPA Guideline 60 gives examples of triggers for an interim ORSA: a significant M&A transaction or portfolio transfer, a major catastrophe loss materially affecting the solvency position, a significant change in the business plan or risk appetite, a material change in the reinsurance programme, or a change in the regulatory capital framework that materially affects SCR calculation. The NCA may also request an interim ORSA as part of supervisory dialogue. For most undertakings the annual ORSA cycle aligns with the business planning cycle in Q3-Q4.