Generative AI unlocks the value of unstructured data in insurance

Generative AI (GenAI) has the potential to reshape the insurance industry more profoundly than any previous wave of automation.

“As an industry, insurance runs on data — but most of it isn’t cleanly structured,” says Łukasz Cempulik, DWH and BI Architect at Striped Giraffe. “We’re talking about thousands of scanned documents, handwritten forms, medical reports, legal correspondence, photos, videos, voice recordings — data that has historically required manual handling.”

With large language models (LLMs) and multimodal AI, insurers can now tap into this ocean of unstructured information to drive faster, smarter, and more consistent decisions — while reducing costs and improving service levels.

Cempulik adds: “GenAI doesn’t just automate existing processes — it introduces an entirely new level of cognitive capability. It can read a 40-page medical record, cross-reference it with actuarial tables, and summarize key risks in seconds. That’s enabling a different scale and depth of insight.”

How exactly are insurers putting these new capabilities to work? Let’s explore the key use cases:

1. Risk Assessment & Underwriting

Instead of sampling limited data points, GenAI reviews full electronic health records (EHRs), lab results, financial disclosures, wearable sensor logs, and even lifestyle indicators. It interprets hundreds of pages of medical records and cross-references them with actuarial data to surface risk signals no human would spot.

Impact: More accurate, objective underwriting — completed in hours, not weeks.

Traditionally, underwriting depended on predefined checklists and summary data. GenAI changes that by consuming full electronic health records (EHRs), diagnostic reports, financial statements, lifestyle surveys, wearable sensor logs, and even social media behavior (where permitted by regulation).

In practice, a GenAI model can:

• Parse hundreds of pages of clinical documents using natural language understanding.
• Extract relevant biomarkers, pre-existing conditions, or risk flags.
• Compare findings with actuarial models and propose a risk score.
• Draft underwriting recommendations with explainable reasoning.

This leads to more accurate and objective risk assessments — and far faster turnaround times.

Real-world example: A leading life insurer in Germany now uses a GenAI-based underwriting assistant that evaluates full medical histories and lab reports. Underwriters report a 60% reduction in time needed for complex case reviews, with improved risk segmentation outcomes.

2. Claims Processing

Claims often involve unstructured content: handwritten notes, PDFs, phone transcripts, photos or videos from property inspections, and even drone footage. GenAI excels at synthesizing such data.

Insurers deploy it to:

• Digitize and categorize scanned forms.
• Generate structured claims summaries based on adjuster narratives and evidence.
• Detect gaps, inconsistencies, or duplicate submissions.
• Auto-suggest decision pathways based on historical resolutions.

Integration with OCR (Optical Character Recognition), computer vision, and document classification tools enhances this pipeline.

Real-world example: A large multiline insurer in the Netherlands has integrated GenAI with its claims triage system. The solution reviews thousands of property damage claims weekly, cross-checking adjuster notes with photographic evidence and flagging inconsistencies. Claim resolution times have dropped by over 40%.

3. Fraud Detection

Fraud schemes often involve subtle patterns across communications, claims, and behavioral anomalies. GenAI brings a semantic lens to fraud detection.

Its capabilities include:

• Analyzing historical cases to uncover fraud indicators across text, image, and voice.
• Identifying linguistic inconsistencies across documents and emails.
• Spotting behavior shifts in claimant communication or submission frequency.
• Collaborating with machine learning models for anomaly detection.

Real-world example: A Scandinavian insurer integrated LLMs into their anti-fraud program. The system surfaced coordinated fraud rings based on similar phrasing in multiple claims. Within six months, the detection rate of suspicious cases increased by 28%.

4. Product Development

In a rapidly shifting market, GenAI enables insurers to innovate faster and more precisely.

It can:

• Analyze customer feedback from call center logs, surveys, and online reviews.
• Benchmark competitor offerings using scraped policy content.
• Generate ideas for new insurance products or dynamic pricing models.
• Simulate customer responses via synthetic personas.

Paired with traditional analytics, this leads to customer-centric innovation and faster go-to-market cycles.

Real-world example: A mid-sized insurer in Central Europe used GenAI to develop micro-insurance products for gig workers and freelancers. By analyzing usage patterns, customer complaints, and social media threads, the company introduced two new offerings that reached profitability within nine months.

5. Customer Service

Perhaps the most visible application, GenAI powers AI assistants that understand context, nuance, and policy complexity.

These systems:

• Retrieve answers from policy documents, prior correspondence, and regulatory guidelines.
• Generate personalized responses across channels.
• Understand sentiment, intent, and urgency to escalate when needed.
• Offer multilingual support for international markets.

GenAI also assists live agents by summarizing conversation history and suggesting next-best actions.

Real-world example: A major health insurer in the US deployed a GenAI-powered chatbot across web and mobile apps. Within weeks, it was handling over 70% of member inquiries without human intervention — maintaining a 92% satisfaction score.

Further application scenarios

Beyond the five areas outlined above, GenAI is beginning to influence:

• Regulatory reporting: Drafting compliance reports and explaining decisions in natural language.
• Broker enablement: Equipping sales reps with AI-generated product comparisons and selling points.
• Policy management: Simplifying coverage changes and policy renewals via conversational interfaces.

Transformation Starts with Data

The success of all of the above use cases hinges on a critical enabler: unstructured data readiness for AI.

“Unlocking GenAI’s full potential depends on how well insurers prepare their data,” emphasizes Cempulik. “You can’t just drop PDFs into a model and expect magic. You need to clean, classify, and annotate your unstructured content. That’s where proper data pipelines and governance come in.”

To prepare unstructured data for GenAI, insurers are increasingly investing in:

• Document digitization and OCR optimization.
• Data labeling and metadata enrichment.
• Knowledge graph integration for contextual understanding.
• AI governance frameworks ensuring explainability and compliance.

Done right, these investments not only improve GenAI performance but also lay the foundation for more transparent, auditable, and secure AI systems.

Conclusion

Generative AI is not a distant promise — it’s already reshaping how insurance works, from the underwriting desk to the customer help line. By unlocking the value in messy, unstructured data, GenAI gives insurers a competitive edge in speed, precision, and personalization.

But that edge won’t come by itself. It requires robust data strategies, interdisciplinary collaboration, and a strong understanding of GenAI’s capabilities and limits. For insurers ready to rethink their approach to information, the rewards are clear: faster insights, better products, and more satisfied customers.

“The real disruption isn’t just AI implementation — it’s what becomes possible when you prepare your data so AI can finally understand it and make use of it,” Cempulik adds.

You may also like:

• E-Book: Digitalization in the financial sector
• Blog article: Top 7 challenges of digitization in the financial sector
• Expertise: AI Readiness