AI underwriting (batch 12 expansion)

Category: AI in insurance · Reviewed by Al Jabbar, Broker · Specialist Risks · Last reviewed 2026-06-10

AI underwriting is the application of artificial intelligence and machine-learning techniques to risk selection, pricing and acceptance decisions across personal, commercial and specialty lines, undertaken by United Kingdom insurers and Lloyd’s syndicates within the governance expectations set out by the Financial Conduct Authority (FCA) and the Prudential Regulation Authority (PRA). This expansion to the earlier AI underwriting entry reflects the post-2023 supervisory environment, particularly the Feedback Statement FS2/23 and the maturing EIOPA AI governance work.

Category: AI in insurance · Aliases: AI-augmented underwriting, AI risk selection · Established: AI underwriting in UK personal lines pricing matured c.2010-2018; specialty / Lloyd’s adoption accelerated 2020-2025 · Related: Algorithmic underwriting, Machine learning underwriting, Insurtech

Definition

AI underwriting denotes the use of statistical learning algorithms — generalised linear models (GLMs), gradient-boosting machines (GBMs), neural networks and, increasingly, large language models — to support or automate the assessment of insurance risks. In the United Kingdom market the term covers a spectrum: from fully automated decisioning in motor and household personal lines, through analyst-in-the-loop triage in commercial lines, to retrieval and summarisation copilots used by Lloyd’s underwriters at the box.

Legal / Regulatory basis

AI underwriting in the UK is governed by a combination of horizontal data protection law, financial-services conduct and prudential rules, and emerging AI-specific guidance:

• FCA Handbook: SYSC 4 (general organisational requirements), SYSC 7 (risk control), PRIN (the Principles for Businesses, including the Consumer Duty cross-cutting rules under PRIN 2A) and ICOBS Chapter 5 (identifying client needs).
• PRA Rulebook: General Organisational Requirements; expectations on model governance for insurers under Solvency II internal-model rules; PRA SS1/23 Model risk management principles for banks (May 2023) is formally addressed to banks but is treated as directional by insurance model-risk teams.
• FCA and PRA Discussion Paper DP5/22, Artificial Intelligence and Machine Learning (October 2022), and the joint Feedback Statement FS2/23 (October 2023), which confirmed that the regulators do not intend, in the immediate term, to make AI-specific rulebook changes but expect firms to apply existing frameworks (SYSC, SS1/23, Consumer Duty) to AI use cases.
• Bank of England / FCA Artificial Intelligence Public-Private Forum (AIPPF) Final Report, October 2022.
• EIOPA Artificial Intelligence Governance Principles (June 2021) and the EIOPA Consultation on the Opinion on AI Governance and Risk Management (February 2025) — applicable directly to firms with EU/EEA operations and treated as informative by UK insurers.
• UK GDPR Article 22 (automated individual decision-making) and Data Protection Act 2018; Equality Act 2010 in respect of indirect discrimination and protected characteristics.
• Consumer Duty (PS22/9): AI underwriting that influences price or availability for retail customers must be capable of being explained at outcome level under the Duty’s consumer-understanding outcome.

How it works in practice

A UK insurer or managing general agent typically embeds AI underwriting in a layered pipeline:

1. Data ingestion — proposal-form data, third-party enrichment (DVLA, Companies House, credit reference, perils, geospatial), and historic claims.
2. Feature engineering — derived variables, with documented rationale and bias review.
3. Model training and validation — hold-out and time-based validation, with independent model-validation sign-off (a SYSC 4 governance expectation reinforced by FS2/23).
4. Production deployment — usually as a pricing or referral engine fed into a policy administration system, with documented overrides for underwriters.
5. Monitoring — for drift, performance, fairness metrics, and Consumer Duty outcomes.
6. Human oversight — the AIPPF Final Report repeatedly stresses meaningful human review, particularly where decisions are adverse to a consumer.

Common variations / Subsequent developments

• GLM-based pricing: still the foundation of UK motor and household; see Generalised linear model (GLM).
• GBM and stacked ensembles: increasingly used as a “lift” layer over the GLM; see Gradient boosting machine (GBM).
• Neural networks: used for unstructured inputs (telematics traces, vehicle images); see Neural network underwriting.
• LLM-assisted underwriting: from 2023, syndicates have piloted retrieval-augmented generation (RAG) over their own underwriting guidelines and historic submissions to accelerate triage. See Large language model (LLM) insurance.
• EIOPA Opinion (2025): codifies proportionality, risk-based governance, fairness testing and explainability obligations.

Example

A Lloyd’s syndicate writing UK mid-market property uses a GBM ensemble to score incoming submissions for desirability against its plan. The model returns a propensity-to-bind, an indicative technical price and the top SHAP contributors. Underwriters retain authority; declines and material rate movements are logged with the model’s contribution recorded. The validation team runs a quarterly fairness review across postcode-derived proxies and produces a model-risk report consistent with the firm’s documented framework, which references DP5/22, FS2/23 and the spirit of PRA SS1/23.

See also

• Algorithmic underwriting
• Machine learning underwriting
• Explainable AI insurance
• AI ethics insurance
• Insurtech

References

• FCA & PRA, DP5/22 — Artificial Intelligence and Machine Learning, October 2022. https://www.fca.org.uk/publications/discussion-papers/dp5-22-artificial-intelligence-machine-learning
• FCA & PRA, FS2/23 — Feedback Statement on AI and Machine Learning, October 2023. https://www.fca.org.uk
• Bank of England & FCA, Artificial Intelligence Public-Private Forum — Final Report, October 2022. https://www.bankofengland.co.uk
• EIOPA, Artificial Intelligence Governance Principles, June 2021. https://www.eiopa.europa.eu
• EIOPA, Consultation Paper on the Opinion on AI Governance and Risk Management, February 2025.
• PRA, SS1/23 — Model risk management principles for banks, May 2023. https://www.bankofengland.co.uk
• FCA Consumer Duty PS22/9, July 2022.
• UK GDPR; Data Protection Act 2018; Equality Act 2010, https://www.legislation.gov.uk

This entry is part of the Apex Insurance Wiki. Last reviewed by Matt Bartlett on 2026-06-10. Next review: 2026-12-10.

Apex Insurance Brokers Limited. Authorised and regulated by the Financial Conduct Authority, FRN 724952. Registered in England and Wales, Companies House 07014570. This entry provides general information about UK insurance concepts and is not regulated advice. Consult your insurance broker on your specific position.