Insurers face hurdles scaling AI across diverse global regulations
Insurance companies must balance the benefits of AI-driven pricing and distribution with complex, varying international rules and the need for human oversight.
The short version
- AI allows for granular risk assessment but can create unfair pricing differences for groups with similar risk levels.
- Third-party AI assistants may shape customer decisions before they ever reach a regulated insurance provider.
- Different markets have different rules for AI transparency, such as Singapore's lifecycle controls and the US NAIC Model Bulletin.
- Insurers must ensure a named executive remains accountable for AI outcomes to maintain trust and regulatory compliance.
Why it matters
For the average consumer, this means insurance prices and claims could be increasingly influenced by AI models that are harder to see. For the industry, it means companies must build robust governance to prove their AI isn't discriminating or operating without human oversight.
Moving from pilots to core operations
Many insurance companies are looking to move past small, local AI proofs of concept toward broader adoption across different products and jurisdictions. This ambition is hitting a wall of reality because the most impactful areas—like pricing, distribution, and claims—are governed by long-standing regulations. As EY reported, firms are now forced to adapt their risk frameworks to keep pace with this shift from testing to core business operations.
The challenge lies in balancing innovation with consumer protection. While these two goals are often seen as competing, sustainable growth depends on the ability to scale new capabilities within trusted boundaries. Management must figure out how to apply established principles to a landscape where decisions are becoming more data-intensive and less explainable.
Insurers also face novel risks as AI is given the authority to recommend or execute actions on a firm's behalf. These risks include a heavier reliance on third parties and the potential for AI to operate under evolving conditions. The goal is to convert these complex journeys into measurable customer and financial outcomes at a large scale.
The complexity of AI-driven pricing
AI enables insurers to analyze more data to draw finer distinctions between different customers. While this can improve pricing accuracy, it can also produce greater differences between people than regulators or consumers might find acceptable. For example, people with similar risk levels might receive different prices based on behavioral traits like their likelihood of switching providers.
The report notes that pricing models might also use factors that appear neutral individually but result in less favorable outcomes for specific groups when combined. These issues are distinct: risk pooling is about affordability, differential pricing is about conduct, and proxy discrimination is about legal compliance. Treating them as a single fairness problem leads to the wrong type of controls.
Boards and management teams need to set clear boundaries for what pricing outcomes are acceptable and define when a result should be escalated. These boundaries only work if the insurer can show exactly how a price was reached. This requires embedding these checks across underwriting, actuarial, and product-governance processes to ensure accountability.
Shifting customer interfaces
A new scenario is emerging where customers begin their search for insurance using third-party AI assistants rather than an insurer's own website. These general-purpose AI interfaces may compare products, narrow down options, or recommend a specific provider before the customer ever reaches a regulated entity. This is particularly relevant for simple non-life retail products where online distribution is already well-established.
Insurers may find it difficult to recover a customer relationship if they lose the initial moment of decision. Because AI assistants might rank products based on commercial relationships or incomplete information, customers may assume these recommendations are neutral. This creates a risk where products that are hard for an AI to find may become less visible, regardless of their actual quality or suitability.
To prepare for this, boards should decide which parts of the customer experience they want to control directly and which they are willing to leave to external platforms. This is especially important for complex decisions, like industrial risk insurance, where physical and hybrid channels still remain more significant than online-only models.
Delegating authority to systems
Every insurer already decides who can make a decision and when to seek approval, but giving AI a role in this process requires deciding what specific authority it should have. An AI application might support an underwriter, recommend a claim handling method, or execute an action without any human approval. Each step toward autonomy increases the consequences of the system getting a decision wrong.
Regulatory requirements for automated decision-making and explainability do not always align across different global markets. For instance, Singapore has proposed guidelines on lifecycle controls for insurance underwriting. In the United States, the 2023 NAIC Model Bulletin sets baselines for governance in 25 jurisdictions to ensure AI does not result in unfair practices.
Insurers must set clear limits on what actions AI can take and identify when human intervention is mandatory. A named executive must remain accountable for the final outcome and have the power to stop the system's use. These boundaries should be proportionate to the customer impact, the data used, and the overall operational significance of the task.
Tracing outcomes across journeys
Insurers typically organize their accountability around specific products, legal entities, or individual models. However, customers only experience the final outcome, not the internal organizational structure. A problem can emerge when a pricing model, an AI assistant, and a claims process all perform as intended separately but create issues when their decisions interact during a single customer journey.
As more teams and third parties contribute to an AI-enabled outcome, it becomes harder to trace who is responsible for the result. Supervisors may eventually look more closely at how these elements connect across the entire customer journey. This approach would support more consistent consumer protection as digital interactions cross traditional institutional boundaries.
Before scaling material journeys, insurers should evaluate four key areas: value, boundary, accountability, and contestability. They must ask if the AI supports or executes a decision, if a named executive is responsible for the third-party chain, and if a consequential outcome can be explained and corrected. Companies that can answer these questions will be better positioned to scale AI with confidence.
What to watch
- How regulators will harmonize rules on 'proxy discrimination' in AI pricing.
- Whether third-party AI assistants will begin to demand direct commercial partnerships with insurers.
Try it
Check if your current insurance provider offers a direct AI chat assistant or if you primarily use third-party comparison sites to find quotes.
Sources and links
- ReportingEY: the article we sourced
First reported by EY. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.