Skip to main content

Controlling the Underwriting Losses: How AI Can Revolutionize Insurance Pricing

Controlling the Underwriting Losses: How AI Can Revolutionize Insurance Pricing

Underwriting is a critical function that determines the financial health of an insurance company. However, underwriting losses has been a persistent challenge, affecting the profitability and sustainability of insurers. Here we explore how good underwriting practices, coupled with AI and ML data-driven effective pricing decisions, can significantly mitigate underwriting losses, and drive better business outcomes.

Predictive Analytics to Improve Underwriting Productivity

The global market of AI in insurance was valued at $4.59 billion in 2022 and is projected to reach $79.86 billion by 2032, growing at a CAGR of 33.06%. Underwriting is considered one of the largest use case segments for AI in insurance. Combined & Loss Ratio improvement using AI ML model has been reported as 3-6 & 2.1-4.2 percentage points respectively. With predictive analytics, underwriting productivity has been estimated to improve by 50%, anti-selection of portfolio improvement 10-15%, additional GWP growth by 3-4%.

Good Underwriting, a Careful Balancing Act

In the complex world of insurance, mastery of underwriting is crucial for insurers to manage risk and protect their finances. Good underwriting involves assessing and managing risks effectively, ensuring that premiums cover expected losses. It involves a careful balancing act, between evaluating potential claims and setting appropriate premiums.

Underwriting losses occur when the premiums collected are insufficient to cover the costs of claims and operating expenses. These losses can adversely impact an insurer's financial health, leading to decreased profitability, diminished market reputation, and even insolvency in severe cases. Efficient pricing plays a crucial role in covering underwriting losses by ensuring that premiums accurately reflect the risks assumed by the insurer.

Natural Catastrophes and Underwriting Losses

In 2023, the Property and Casualty insurance sector faced significant challenges. According to Fitch Ratings, Europe's leading reinsurers like Hannover Re, Munich Re, SCOR, and Swiss Re were affected by weaker underwriting and investment losses. Recent reports highlight significant financial impacts on the insurance industry due to natural catastrophes and underwriting losses.

According to Munich Re, Europe has faced a series of billion-euro market losses from natural catastrophes, emphasizing the need for disciplined underwriting. SwissRe reported that underwriting losses in the first half of 2023 reached USD 22 billion, resulting in a net income of just USD 2 billion, despite increased investment earnings.

In the US, the property and casualty (P&C) sector experienced a net underwriting loss of $32.2 billion in the first nine months of 2023, marking a $7.6 billion increase from the previous year. AM Best’s First Look Report noted a decline in underwriting performance, with the industry’s combined ratio deteriorating to 103.4.

Effective risk assessment and price determination are key factors that contribute to good underwriting practice. This involves:

  • Identifying
  • Rejecting risks that are not favourable carriers, thus improving underwriting profits.
  • Mitigating adverse selection by attracting a balanced proportion of high-risk policyholders
  • Maintaining financial stability by ensuring sufficient funds to pay claims & cover operating costs.
  • Setting premiums that adequately cover expected losses and associated expenses.
  • Determining premiums based on risk characteristics to account for potentially large losses.
  • Adapting to changing market conditions and evolving risk factors

Sustained Underwriting Losses Adversely Impact the Insurer’s Financial Stability & Profitability

Underwriting losses have a significant impact on the performance of insurance companies.

Reinsurers may increase premiums or impose strict terms when underwriting losses occur, adding to the insurer’s overall cost burden. These losses may limit an insurer’s capacity to underwrite new business or expand into new markets, hindering growth opportunities. Recover from underwriting losses, insurers may need to raise premium rates, potentially leading to customer dissatisfaction and loss of market share.

Regulators may increase scrutiny on insurers experiencing underwriting losses, potentially leading to increased regulatory requirements or interventions. If these losses persist, credit rating agencies may downgrade the insurers credit rating, affecting the company’s ability to borrow and increasing the cost of capital. Consistent underwriting losses can impact and harm an insurer’s reputation in the market, making it less attractive to both policyholders and distribution partners.

Role of AI & ML for Effective Data-driven Risk Evaluation

Today, most insurance companies’ premiums are using *Generalized Linear Models (GLM). These models factor in variables related to both the party purchasing the insurance and the object being insured. *Frequency model predicts the number of claims while a *severity model estimates the average amount of one claim based on limited factors. The base rate is then determined and multiplied by other rating factors such as age, marital status, gender, deductible, limit etc.

Core insurance platforms use Rate Books with rate table spreadsheets imported for each country and state. While these rating engines serve their purpose, they lack support for dynamic and data-driven pricing calculations.

As rating is a critical component in the policy lifecycle, most of the modern Policy Admin Systems (PAS) are designed to integrate with external third-party rating systems that support real-time risk assessment with machine learning modelling.

Generalised linear models (GLMs) are a means of modelling the relationship between a variable whose outcome we wish to predict and one or more explanatory variables. Builds a linear relationship between the response and predictors, by using a link function.

Frequency model predicts number of claims that are expected to occur within a given period for a specific policyholder, group of policyholders, or portfolio of insurance policies. Frequency refers to the rate at which insured events, such as accidents, or property damage, are expected to occur.

Severity model predicts average amount of single claim, also known as claim severity or claim size. Analyses historical claims data and factors that influence size of claim like type of loss, coverage limit, deductibles, demographic factors etc.

Machine Learning for Insurance Pricing Optimization

Relevant data sources are used to gather information such as customer information, financial records, past claims data, number and severity of accidents, geospatial data, GPS driving data, accident, theft, catastrophic data for areas from federal and state organizations. Market trends and any other relevant information is also collected. This data is then cleansed to remove any inconsistencies, errors, or missing values.

AI/ML algorithms like logistic regression, decision trees, random forests, gradient boosting machines, naive bayes, neural networks etc. are used to perform risk assessments. These algorithms interpret the importance of each feature in a model, contributing to the model's output. Risk assessment score is generated for each policyholder based on the outputs of the AI/ML models.

Relevant traditional actuarial factors suitable for modelling such as historical claims data, loss ratios, expenses, profit margins, industry standards, demand, competitor pricing, behaviour preferences, and purchase history are selected. These factors are combined with risk assessment scores into a single structured dataset using mathematical functions or algorithms. The dataset comprises rows representing individual policies or insured objects and columns representing predictive variables, including both traditional actuarial factors and AI/ML risk assessment scores.

This integrated dataset involving weightage of varied factors is fed into predictive models like fully connected neural networks. These models identify patterns, relationships, and correlations between the several factors and insurance risk. Output determines the accurate rate and customer segments (low/ medium/ high), as well as conversion rates based on customer behaviour and purchase history.

Adjustment factors such as discounts for low-risk, surcharges for high-risk, deductibles and expense loading percentages covering administrative/ operational expense to specific risk profiles are then applied to determine the final premiums.

Additional adjustment gets auto-applied based on catastrophic modelling and reinsurance costs for risk exposed to catastrophic events. Model can be further calibrated based on risk appetite and expert judgment to ensure it reflects the relationship between risk assessment scores, actuarial factors, and insurance premiums.

Why Coforge

At Coforge, we empower leading insurance companies with innovative technology solutions to achieve their business goals and build resilience. Our team of 200+ Certified Data Scientists serves over 45 global customers, delivering 220+ solutions that have demonstrated:

  • 95% increased efficiency
  • 99% improved data accuracy
  • 94% reduced processing time
  • 45% reduction in fraud cases
  • 5% overall error rate

We tailor our solutions to meet the unique needs of insurers worldwide, leveraging significant partnerships with:

  • AI Hyperscalers: Azure, AWS, Google Cloud
  • AI-Infused Data Platforms: Snowflake, Databricks, Pega, Salesforce
  • AI ML Ops: Dataiku, DataRobot
  • Academia: MIT, Penn

These collaborations enhance our AI capabilities and client value.

Our Quasar GenAI platform builds enterprise AI capabilities with solutions and accelerators for pricing accuracy, offering access across Hyper Scalars and Open Source LLM. As a premier Azure OpenAI partner, our GenAI solutions are marketplace-listed.

With deep AI/ML expertise, we deliver comprehensive risk assessment and pricing solutions for underwriting, integrating Gen AI, machine learning, and predictive analytics.

For more information, visit or e-mail:

Let’s engage