Commercial insurance underwriting is at a critical turning point. With revenues climbing 8% annually over the past five years and a remarkable combined ratio of 91%, the numbers look promising. But the industry isn’t free from challenges. Insurers are under increasing pressure to reduce costs, enhance risk assessment, and achieve profitable growth. The root of the problem? Data.
On the Unstructured Unlocked podcast, Laura Drabik, Chief Evangelist at Guidewire, explored a compelling issue plaguing the industry. Despite having access to unprecedented amounts of data, insurance professionals remain buried under a mountain of unstructured, fragmented information. During the episode, Laura discussed how generative AI (GenAI) and agentic AI will shape underwriting’s future by streamlining operations, bolstering efficiency, and improving decision-making across the board.
This post highlights the key insights from Laura’s conversation with Tom Wilde, CEO of Indico Data, as they dove into the innovative ways artificial intelligence is revolutionizing commercial insurance.
Listen to the full podcast episode here: Unstructured Unlocked Podcast
The structured data problem in underwriting
Insurance is, at its core, a decision-making industry. Underwriters must weigh risks, evaluate pricing, and make judgments to achieve solid profit margins. But as Tom Wilde noted during the podcast, these decisions hinge on structured, high-quality data. Unfortunately, much of the available information remains unstructured.
Laura mentions,
“Only 25% of broker submissions become written policies. Up to 60% aren’t even reviewed. Not due to a lack of opportunity, but because underwriters are buried in fragmented, unstructured data.”
Why is this happening? The reality is that new broker submissions are arriving at a rate that carriers can’t keep up with. Submissions often land in centralized inboxes or with individual underwriters, coupled with a hodgepodge of email bodies, PDFs, and spreadsheets. Extracting key insights from these formats is a tedious and manual process. As a result, only a small portion of submissions are considered for review.
But it isn’t just about the submission process. Underwriters touch as many as 15 systems daily, according to Laura. From internal databases to risk modeling tools, the ecosystem is fragmented, creating inefficiencies and missed opportunities.
How generative AI and agentic AI can transform underwriting
AI offers the tools needed to tackle the data problem head-on. Generative AI and agentic AI are at the forefront of this transformation.
Generative AI’s role in underwriting
Laura identified generative AI as a crucial tool for summarization, context extraction, and decision support. It can process large volumes of unstructured data and turn it into actionable insights, ultimately making underwriters more effective. For example:
- Improved triage: GenAI can sift through broker submissions and quickly flag the most promising opportunities, reducing time wasted on low-priority submissions.
- Better risk assessment: By combining broker-submitted data with external and internal sources, generative AI provides a clearer picture of overall risk.
However, as Tom Wilde explained, generative AI isn’t suited for everything. While it excels at interpretive tasks, such as summarizing underwriting guidelines, it struggles with deterministic decisions, where data accuracy is non-negotiable. That’s where agentic AI steps in.
Related Content: How to implement underwriting automation without disrupting your existing workflows
What agentic AI brings to the table
Agentic AI is another exciting development in the AI space. Designed to act autonomously, it can make decisions and take actions to meet defined business goals. The technology allows insurers to:
- Triage and pre-clear submissions to ensure only viable risks reach underwriters.
- Orchestrate various data sources and combine them into structured frameworks insurers can trust.
- Recommend next best actions, serving as a copilot for underwriters.
However, as both Laura and Tom emphasized, human oversight remains crucial. Most underwriting decisions are deterministic in nature, requiring accurate, structured data. Agentic AI thrives when paired with human supervision to validate and fine-tune outputs systematically.
Tom explained it best:
“Generative and agentic AI solutions serve as a ‘bionic arm’ for underwriters—not replacing the human element but enabling professionals to scale their expertise and focus on higher-value tasks.”
Redefining underwriting efficiency with AI
How can AI specifically elevate the underwriting process? Laura and Tom identified several strategies for insurers to modernize underwriting practices through AI:
Focus on quality over quantity
Increasing submission throughput isn’t always the solution. Rather, insurers should prioritize techniques that improve triage accuracy and submission quality. AI allows carriers to identify incomplete submissions early, ensuring underwriters focus on decision-ready risks.
Use AI to integrate third-party data
Underwriting decisions don’t solely rely on broker-submitted information. Only 40–50% of data required for risk assessment is typically provided; the rest is sourced from internal or third-party systems. AI platforms can gather external data efficiently, from weather patterns to financial records, to enrich risk modeling.
Human-in-the-loop design
AI implementation shouldn’t aim to eliminate underwriters. Instead, human-in-the-loop systems ensure professionals remain central to the process, overseeing critical decisions while making use of AI-generated insights.
Building trust in AI systems
For any AI implementation to succeed, trust must be built between underwriters and the tools they use. AI needs to earn confidence by producing reliable, measurable, and traceable outcomes.
Establish “ground truth”
A common challenge is establishing clear metrics for success, also known as “ground truth.” Organizations must define benchmark performance standards to measure whether AI-based systems are effectively streamlining workflows, improving data accuracy, and creating meaningful ROI.
Transparency is key
AI must be transparent and auditable. Every data point and decision derived by AI tools should be traceable to ensure accuracy and compliance with regulations.
Start small and scale
Laura and Tom agreed that AI adoption should begin incrementally. Implementing AI at scale without proper planning risks poor performance and inefficiencies. By starting with specific use cases (like triage or pre-clearance), insurers can refine AI outputs before wider rollouts.
Related Content: Beyond document extraction: how AI is reshaping insurance decisioning
Smarter technology, smarter decisions
The insurance industry has long been rooted in tradition, but innovation is no longer optional. Emerging technologies like generative AI and agentic AI are poised to transform underwriting, enabling insurers to triage faster, assess risks more accurately, and achieve profitable growth in transitioning markets.
Laura summed it up perfectly:
“The future of underwriting isn’t just digital. It’s about making smarter decisions, faster—with the right data at the right time.”
By leveraging AI, carriers can create more efficient decision-making supply chains, outperform competitors, and improve customer satisfaction. But the key lies in thoughtful implementation, robust governance, and continuous improvement.
To learn more and keep up with the latest trends in AI, data, and insurance, make sure to subscribe to the Unstructured Unlocked podcast on your favorite platforms, including:
Subscribe to our LinkedIn newsletter.
Frequently asked questions
- What types of risks or limitations could arise when using agentic AI in underwriting without sufficient human oversight? Without human oversight, agentic AI could make decisions based on incomplete, biased, or inaccurate data inputs, leading to poor underwriting outcomes or compliance issues. There’s also the risk of over-automation, where nuanced judgments—like interpreting ambiguous policy details or market-specific risks—are mishandled by AI systems lacking contextual awareness.
- How do underwriters build trust in AI systems beyond transparency and auditability? Trust is also built through consistent performance, peer adoption, and hands-on training. Underwriters need to see tangible benefits—such as reduced processing times or improved risk accuracy—and must be included in the design and feedback loops of AI tools. Empowering them to understand, challenge, and refine AI recommendations builds ownership and credibility over time.
- What are the infrastructure or integration challenges insurers face when implementing generative and agentic AI? Many insurers operate on legacy systems that aren’t compatible with modern AI tools. Integrating AI often requires significant data cleaning, re-architecture of workflows, and alignment across fragmented systems. Insurers also need strong data governance frameworks and APIs to enable smooth AI adoption without disrupting compliance or operational continuity.
