Causal AI For Insurance
Insurance carriers are leveraging AI to manage increased competition, tighter regulations, complex claims, fraudulent behavior, and improve the overall customer experience. However, current machine learning systems fail to deliver.
Causal AI identifies the true cause-and-effect relationships that govern insurance risk and enables insurers to hedge themselves against the risk of uncertain financial losses.
Looking Beyond Tomorrow
- Insurtechs expansion
- Cyber security risk
- IoTs changing what and how we know it
- Digital Transformation
- Fraud and Risk Management
- Regulatory changes
Solutions
Move from ROI to ROX and embed your clients by using Causal AI to better know what drives your clients, their needs and how to deliver solutions most effectively. Acquiring new customers is 5 times more costly than retaining existing ones. Insurers rely on machine learning to proactively identify the policyholders most likely to not to renew. These high-churn risk policyholders are then targeted with retention offers by the marketing team. A majority of those identified would leave despite the retention activity.
The goal is not just to target the policyholders most likely to churn, but instead to target the policyholders that the incentive will cause to stay. However, the greatest cause of churn is something that insurers can address – that is – a poor claims experience.
Insurers can use causal AI to:
- Predict the likelihood of churn of a policyholder
- Understand the causal drivers of policy churn and the impact of external factors on the churn rate
- How does a 5% increase in average customer satisfaction affect the policy churn rate?
- What if our competitor starts offering greater coverage at the same premium price. How would that affect the churn rate?
- Shape their business outcomes
- What if we offer a 10% discount on premium prices to the policyholders on which the discount would yield the greatest ROI. How would that affect the churn rate?
Every 5 minutes an insurance fraud is uncovered. Bad actors, identity theft, and scams cost the insurance industry more than £3.3 million every day.
Insurance providers turn to AI to mitigate losses by identifying potentially fraudulent claims for investigation. Current ML systems pick up spurious correlations that increase the frequency of false alerts and negatively impact customer satisfaction. Causal AI zeros in on the factors that truly drive fraudulent claims thereby balancing the savings from loss prevention with the cost of false alerts.
Insurers can use causal AI to:
- Predict the likelihood of a claim being fraudulent
- Understand the causal factors behind fraudulent claims and the impact of external factors on the losses due to fraud
- What if unemployment rises by 5%. How would that affect the total number of fraudulent claims?
- Shape their business outcomes
- What if we increase the fraud threshold for manual review of auto-insurance claims by 10%. How does that affect customer satisfaction levels and in turn the total PnL of the auto-insurance portfolio?
Insurers today have access to new data sources such as telematics that are used in auto insurance and fitness data in health insurance – all of which help to underwrite risks more accurately. Insurers are migrating away from age-old GLMs to state-of-the-art Machine Learning methods for underwriting. However, these methods fall short as conventional machine learning algorithms produce spurious and misleading correlations. They also absorb biases that are implicit in the training data.
Causal models contain a transparent qualitative component that describes all cause-and-effect relationships in the data, and so there are no problems of trust, fragility or fair washing.
Insurers can use causal AI to:
- Predict expected claim amount for a given customer
- Understand the impact of external factors on the Insurance portfolio
- What if the total number of COVID cases rises by 10%. How would that impact the average loss ratio of the Health Insurance portfolio?
- Shape their business outcomes
- What if we offer a free wellness program such as smoking cessation classes to health insurance policyholders. How does that impact the average loss ratio of the health insurance portfolio?
AI is being embraced but is it effective or falling short? Machine learning tends to overfit and underdeliver. Typical ML solutions are black boxes that don’t meet regulators’ transparency hurdles. Incoming regulations in the EU demand explainability for higher risk systems, with fines of up to 4% of annual revenue for non-compliance.
“There is a gap between explainability in practice and the goal of transparency”, according to research conducted by the Partnership on AI.
Causal models contain a transparent qualitative component that describes all cause-and-effect relationships in the data, and so there are no problems of trust, fragility, or fair washing. Causal models don’t require another model to approximate them.
Most significantly, causal explanations are generated earlier in the AI pipeline, before models are built. Practitioners can restrict models with their domain knowledge and impose fairness criteria prior to full model building. This enables safer, more compliant and more controlled applications. In contrast to standard XAI, Causal AI provides ante hoc (“before the event”) explainability that is less risky and less resource hungry and critically meets the demands of regulators and users alike.
