This article is for informational purposes only and does not constitute financial or legal advice. You should consult with a qualified professional before making any decisions about the use of AI in your firm.
How Should We Review AI Underwriting Models Against Consumer Duty?
Before using AI to support underwriting, an insurer should test how the model affects eligibility, price and customer understanding. Track outcomes and exceptions across customer groups, and set a route to challenge or pause the system when harm appears.
For the underwriting activity, establish the applicable Consumer Duty provisions with the responsible reviewer. Link the pricing and eligibility tests to the product, customers and confirmed requirements; do not describe automated pricing as automated compliance.
Proxy and subgroup testing. Examine whether inputs or outcomes differ across relevant customer groups, with the legal and compliance team deciding which comparisons are appropriate. Record observed results and investigate disparities; this example does not report an FCA finding about particular firms.
Include data quality, proxy inputs and subgroup outcomes in the underwriting assessment. Record the observed results; no finding about particular firms or systems is established by this example.
- Proxy and subgroup testing. Examine whether inputs or outcomes differ across relevant customer groups, with the legal and compliance team deciding which comparisons are appropriate. Record observed results and investigate disparities; this example does not report an FCA finding about particular firms.
- The "Poverty Premium": AI-driven hyper-personalisation can lead to disadvantaged customers being charged more or being deemed "uninsurable," effectively locking them out of the market.
These are not unfortunate side effects; they are foreseeable harms. Under the Consumer Duty, you have a duty to avoid them. If your AI model is producing these outcomes, you are in breach.
Stress-Testing Your AI Against the Four Consumer Duty Outcomes
Review how your AI underwriting model affects each of the four Consumer Duty outcomes below:
| Consumer Duty Outcome |
How Your AI Underwriting Model Can Fail |
| 1. Products & Services |
Your model could define a target market so narrowly that it systematically excludes certain groups of customers, even if they would be suitable for the product. |
| 2. Price & Value |
Your algorithmic pricing could create a "loyalty penalty" or charge different prices for the same risk, leading to a failure to provide fair value. |
| 3. Consumer Understanding |
If you cannot explain in simple terms why your AI has declined an application or charged a certain premium, you are failing the consumer understanding outcome. |
| 4. Consumer Support |
If a customer queries a decision made by your AI, and your staff are unable to provide a clear and coherent explanation, you are not providing adequate support. |
A Practical Framework for Compliant AI Underwriting
Define who tests the model, reviews its decisions and responds when customer outcomes cause concern. Record those responsibilities in the governance framework.
- Conduct Pre-Deployment Bias Audits: Before deploying a new model, test its outputs across demographic groups for bias and discriminatory patterns.
- Scrutinise Your Data: Review the model's internal and third-party data sources. Document how you identify and mitigate biases in those sources.
- Implement Meaningful Human Oversight: Assign a reviewer who can challenge and override AI decisions, particularly in complex cases, cases involving vulnerable customers and declined applications. Give that reviewer authority to intervene.
- Prioritise Explainability: Explain the factors that drive the model's decisions. If the model is a "black box", explain the safeguards and controls in place to prevent unfair outcomes.
- Monitor Continuously: Monitor the model's performance in the live environment for "model drift" and emerging biases.
Review customer outcomes and keep evidence
Use current primary supervisory material when assessing the proposed insurance use. Record the relevant source and applicability decision instead of inferring enforcement intent or political motivation.
Review the model's impact on customers alongside its technical performance.
Keep evidence of fairness, transparency and customer outcomes so the firm can review and respond to compliance risks.
Take the Next Step
The AI Audit is a starting point for reviewing AI readiness, governance gaps and risks in current tooling. Use that review to decide what the firm needs to address.
Book a Discovery Call → or learn more about the AI Audit.
Need help assessing an AI workflow and its unresolved risks?
Map where AI can improve decisions, service and operations while keeping Consumer Duty, SM&CR, model-risk and client-data controls visible.
