The transformative power of AI in Underwriting: Exploring practical applications

With experience both as an actuary and underwriter, Risa Ryan has spent more than 25 years in the insurance industry. In this blog, she shares her thoughts on how AI and ML will address key challenges for underwriters.

The insurance industry stands at a pivotal juncture, with AI and Machine Learning transforming commercial underwriting. This transformation is accelerating rapidly, with GenAI adoption among P&C insurers jumping from 18% to 63% in just one year.

While AI enables better risk assessment and addresses inefficiencies in underwriting workflows, a significant implementation gap exists. Research shows only 4% of insurers have fully deployed AI capabilities despite 42% running pilots, highlighting the challenges in moving from experimentation to production.

This article will delve into details of how machine learning and AI can provide robust solutions to some of the industry's most pressing challenges, and aims to explain the practical applications of how and where this exciting new tech will sit in the underwriting workflow in 2025.

What are the current challenges in underwriting?

The insurance industry has often fallen behind other financial sectors in embracing and adopting new technologies, leading to inefficiencies and missed opportunities, including:

1. Manual data entry and administrative burden

Rekeying in insurance is commonplace, with our 2024 State of Pricing Report finding that underwriters spend on average 3 hours each day manually entering data. Accenture research shows approximately 40% of commercial lines underwriters' time is spent on non-core administrative tasks, including redundant data inputs and rekeying.

This statistic has remained consistent into 2025, with recent industry analysis confirming underwriters still spend nearly half their time normalizing PDFs, spreadsheets, and manually entering data. This process is prone to inaccuracies with resulting errors cascading into pricing discrepancies and affecting the insurer's profitability and risk management.

The challenge intensifies as the industry faces a wave of retirements that some call the "silver tsunami": over the next 15 years, 50% of the current insurance workforce is expected to retire, leaving more than 400,000 positions unfilled. Meanwhile, incoming talent expects modern tools, making efficient workflows critical for both knowledge transfer and capacity management.

2. Fragmented data sources and system integration challenges

Underwriters often need to access multiple sources to gather necessary information for proper risk assessment. This fragmentation creates inefficiencies and potential data gaps, restricting underwriters' ability to make informed pricing decisions.

The technical infrastructure challenge is severe: 65% of carriers still operate on pre-cloud legacy systems, with 80% experiencing cross-system data harmonization difficulties when attempting to integrate new technologies into existing underwriting workflows.

3. Inconsistent risk assessment across underwriting teams

The effectiveness of risk assessment varies significantly among underwriters based on their individual willingness to explore multiple data sources, their depth of research, and underlying expertise levels. This variance introduces noise and bias into the risk assessment and pricing process.

McKinsey research shows that 71% of insurers in the top quartile by net combined ratio remained there over a decade, while only 18% of middle-performer insurers moved to the top quartile, suggesting that consistent underwriting excellence provides sustained competitive advantage.

4. Unreliable claims and underwriting feedback loops

The feedback loop between claims and underwriting is often inconsistent, limiting teams' ability to learn from past claims and adjust underwriting practices accordingly. This challenge has intensified as the industry posts exceptional performance—the U.S. commercial P&C insurance industry achieved a 96.5% net combined ratio in 2024, the best underwriting results in over a decade, according to S&P Global Market Intelligence analysis.

However, rating agencies with access to regulatory filing data have not isolated AI's specific contribution to these improvements, representing a critical gap in publicly available ROI metrics.

How AI provides solutions to underwriting problems

AI technology offers several innovative solutions to these challenges, particularly through the analysis of unstructured data. From automating manual processes to enhancing risk assessment precision, these capabilities are revolutionizing how insurers operate.

Below are six ways AI provides solutions to underwriting problems:

1. Enhance data entry with Natural Language Processing (NLP)

AI-powered ingestion tools extract data from complex submissions instantly, eliminating the tedious data preparation that consumes underwriter time. Research shows 64% of insurance AI applications focus on document processing, with modern platforms extending capabilities well beyond basic OCR.

The most effective ingestion solutions share several characteristics: they connect directly to pricing and rating systems, allowing extracted data to flow into models without rekeying. They capture submission data as a durable asset, enriching pricing tools with factors underwriters otherwise wouldn't have time to enter. And critically, they maintain human control through review stages, editing capabilities, and confidence scoring, ensuring underwriters stay in command of decision-making rather than blindly accepting AI outputs. This combination of speed and oversight transforms submission intake from a bottleneck into a competitive advantage.

2. Streamline data access with image recognition

AI can significantly reduce the need for underwriters to access multiple data sources by leveraging automatic image recognition. According to DataIntelo research, the aerial imagery market for insurance applications is experiencing significant growth, with insurers increasingly leveraging high-resolution aerial imagery to enhance the accuracy of property assessments and streamline claims management.

For instance, AI-driven predictive models can analyze images to determine the number of stories in a commercial building, roof type, roof quality (including standing water presence), and building age. Specific documented applications include:

• California Wildfire Risk Assessment: Insurers use aerial AI to track brush growth around California homes for refining wildfire risk models and providing more accurate coverage pricing through real-time monitoring of vegetation conditions

• Flood Risk Zone Refinement: AI-driven aerial data helps insurers refine flood-prone zones by discovering urban development patterns that increased runoff risk, leading to adjusted policy coverage

• Property Feature Verification: Insurers cross-check property features with AI-generated insights to detect undisclosed risks for greater accuracy and transparency in underwriting

Commercial platforms like LexisNexis Flyreel demonstrate operational deployment, turning regular mobile devices into AI-enabled property risk assessment tools that create 3D models in less than 10 minutes while collecting essential property insights tailored to specific risks and guidelines. According to the research, Flyreel for Commercial "helps guide busy small business owners through a simple inspection experience, collecting essential property insights tailored to your risks and guidelines."

3. Ensure consistent risk assessment

AI brings a standardized approach to risk assessment, ensuring consistency across the underwriting process. By employing uniform AI-driven methodologies, disparities caused by individual underwriters' experience levels or research thoroughness are eliminated.

However, implementation requires careful balance. More sophisticated AI models often deliver better predictive performance, but their complexity can make regulatory explanation difficult. Research indicates 55% of organizations face challenges explaining AI model decisions to regulators, highlighting the tension between model sophistication and interpretability requirements.

The regulatory framework has matured significantly, with 24 states adopting the NAIC Model Bulletin on the Use of Artificial Intelligence, establishing guidelines for responsible AI use with controls proportionate to consumer harm potential.

4. Improve feedback loops with claims analysis

AI analyzes claims data to identify loss causes, attorney involvement, and claimant sentiment, enabling underwriters to make more informed decisions based on historical patterns. The National Insurance Crime Bureau enhanced fraud detection in 2024, increasing report production by 61% and processing 180,508 questionable claims.

Integrated portfolio intelligence platforms transform historical reporting into forward-looking strategy through real-time tracking and scenario modeling. With 40% of underwriters' time spent on administrative tasks, AI solutions targeting data extraction (64% adoption) and workflow automation (60% adoption) free underwriters to focus on strategic activities while optimizing based on actual results.

5. Fraud detection and prevention

AI can analyze unstructured data from multiple sources to automatically detect fraudulent claims and submissions. According to CLARA Analytics, their AI-based fraud detection for workers' compensation claims has consistently delivered ROI exceeding 500% over the past 10 years for commercial insurance customers.

Looking forward, Deloitte projects that P&C insurers could reduce fraudulent claims and save between $80 billion and $160 billion by 2032 by implementing AI-driven technologies across the claims lifecycle and integrating real-time analysis from multiple modalities including text, images, audio, and video.

High-resolution drone photography also helps in quickly identifying property damage after storms or catastrophic events, aiding in accurate and timely claims processing while reducing investigation costs.

6. Sentiment analysis for enhanced customer insight

Understanding the sentiment of claimants and submission communications can help identify customer service issues and potential influence on claim outcomes. This analysis can lead to proactive measures to improve customer satisfaction and claim resolution.

Advanced AI systems can process multiple data streams simultaneously, analyzing text communications, voice patterns, and behavioral indicators to provide comprehensive risk assessment that extends beyond traditional financial metrics.

Unlocking AI use cases and their benefits

The transformation timeline reveals both significant opportunities and implementation realities. In the short term, AI enhances efficiency by streamlining processes, helping to reduce expense ratios. Survey data indicates that 74% of insurers rolled out new underwriting tools in 2024, with 92% upskilling underwriting teams in data analytics and automation.

However, a critical implementation gap persists. While 42% of P&C insurers are piloting machine learning and GenAI-assisted capabilities, only 4% have fully deployed these technologies in production—revealing a 10x gap between experimentation and operational deployment.

The barriers include:

• Skills and resource constraints (52% of respondents)

• Data quality challenges (40% of respondents)

• Regulatory compliance hurdles (36% of respondents)

In the long term, insurers that successfully navigate these implementation challenges will likely achieve lower combined ratios compared to their counterparts, providing a critical competitive edge. McKinsey research emphasizes that to create lasting business value from AI, insurers need to set bold, enterprise-wide visions and fundamentally rewire how they operate across underwriting, claims, distribution, and customer service.

For insurers looking to move AI from pilot to production, hyperexponential's platform connects submission triage, pricing, and portfolio intelligence in a single workflow, helping underwriting teams close the gap between experimentation and operational deployment.