The insurance industry is on the cusp of a major transformation driven by artificial intelligence (AI) and data analytics. Nowhere is this more apparent than in the critical function of underwriting.
Underwriting is the process of evaluating risk and determining appropriate premiums to accept that risk. It relies heavily on collecting and analyzing data from a wide variety of sources. AI and machine learning are perfectly suited to streamline these data-intensive underwriting workflows.
In 2023, we will see even greater adoption of AI in underwriting as insurers seek to leverage data for smarter, faster underwriting. Let‘s examine how AI is revolutionizing underwriting and what the future looks like.
The Underwriting Bottleneck
Insurance underwriting involves extensive information gathering and risk assessment before quotes can be provided to applicants. Data is collected from multiple sources such as medical records, motor vehicle records, credit reports, geospatial data, IoT sensor data, and more.
Underwriters must pore through all this unstructured data – often still in document form – to identify risk factors and make decisions. As an expert in data extraction, I‘ve seen first-hand the challenges of extracting insights from scanned handwritten forms, PDFs, and images. Important information is essentially trapped on paper.
This manual document-driven process is slow, prone to human error, and prevents underwriters from assessing as much data as they could with automated techniques.
The result is a bottleneck that increases underwriting costs and timelines. Applicants experience long wait times while underwriters are overwhelmed with repetitive administrative tasks. According to McKinsey, 70% of underwriter time is spent on manual data processing.
AI Automates and Augments Underwriting
AI and machine learning provide solutions that both automate repetitive underwriting tasks and augment human underwriters‘ capabilities.
On the automation side, optical character recognition extracts text from scanned documents and forms while natural language processing classifies and extracts insights from text.
For example, Ant Financial‘s imaging recognition technology can extract handwritten and printed text from complex insurance forms 5x faster than humans with 98% accuracy. This reduces document processing times from weeks to days.
Robotic process automation bots can gather data from online sources and populate forms. Munich Re built over 300 software bots handling 1 million transactions per year, saving 85K+ underwriter hours.
This frees up underwriter time from routine paperwork to focus on risk assessment and judgement. It also reduces delays for applicants. One study by Celent found AI could reduce policy application turnaround times by up to 80%.
For augmentation, machine learning algorithms can process far more data than humans to uncover trends and risk factors. Models trained on historical underwriting data learn to predict risks and suitable pricing for new applicants.
According to Capgemini, AI techniques can analyze 50x more data than human underwriters when assessing risk. This provides a complete view of the applicant across all data sources.
Underwriters then leverage these AI insights to make faster, data-backed decisions. With a complete risk picture, they can offer quotes faster and with more confidence. The ability of AI to synthesize insights across disparate data sources is extremely valuable.
Better Risk Assessment with Big Data
One major advantage of AI in underwriting is the ability to assess risk better by utilizing more data sources, including newer alternative data types.
For example, property and casualty insurers can tap into IoT data like smart home device usage when underwriting homeowners insurance. Auto insurers can analyze driver behavior through telematics. Life insurers have begun testing prediction models using fitness tracker data.
In one case, John Hancock was able to use a machine learning model to assess 3 million more data points than human underwriters when evaluating applicants. This increased predictive accuracy by 15%.
Combining structured insurance data with unstructured data from documents, imagery, sensor streams, and more provides a 360-degree view of risk. Deep learning techniques uncover correlations that humans would likely miss.
According to consulting firm McKinsey, AI techniques can lower loss ratios by up to 30% in some lines by better predicting risk.
More Accurate and Profitable Pricing
Better risk assessment powers more accurate premium pricing tuned to the unique risk profile of each applicant. Granular insights allow underwriters to avoid overpricing or underpricing risk.
According to PwC, AI-based pricing can improve premium accuracy by up to 20%. More accurately priced policies mean expanded risk pools and fewer missed approvals or fraudulent claims.
AI also enables insurers to dynamically adjust pricing models in real-time based on the latest data. This keeps premiums competitive and optimized even as new loss patterns emerge.
The result is pricing that expands the risk pool through improved accuracy while maintaining profitability. Applicants get quotes tailored to them rather than broad generalizations.
Per Deloitte, AI adoption could increase life insurance profit margins by up to 25% over time by enabling insurers to right-price policies.
Implementation Challenges
While promising, AI underwriting faces challenges around data, transparency, and change management.
Many insurers lack integrated centralized data platforms to feed AI systems. Cleaning and normalizing disparate datasets can be arduous. Outdated IT systems aren‘t built for big data analytics.
Explainable and transparent AI models are needed to build trust and avoid bias, but can be difficult to achieve with some techniques like neural networks.
Transitioning underwriting workflows to AI requires organizational change management and user training that needs careful planning. Underwriter buy-in is critical.
Overcoming these challenges is key to realizing the full benefits of AI. Insurers are increasingly investing in modernizing their analytics infrastructure and skillsets.
The Future of Underwriting is AI
The success insurers are seeing today only scratches the surface of what‘s possible with AI and underwriting. As data sources proliferate and models improve, AI will drive ever higher efficiency and risk insight gains.
Within a few years, real-time risk scoring fed by connected IoT devices could enable dynamic underwriting and pricing. Claims could trigger automatic underwriting reviews and premium adjustments.
AI will not replace human underwriters entirely. But it will dramatically change their day-to-day work. Underwriters will spend less time processing documents and more time interfacing with AI tools to make strategic decisions.
Adopters of AI underwriting stand to gain a competitive edge through reduced costs, improved risk assessment, and pricing agility. Per IDC, insurers could see up to a 16% reduction in combined ratios through AI adoption by 2022.
AI-powered underwriting will soon become mandatory to remain relevant. The data-driven future of underwriting is coming fast. Insurers need to begin developing the technical capabilities and data infrastructure now to harness its full potential.