Commercial insurance underwriting is undergoing major change. With industry revenues climbing 8% annually over the past five years and a combined ratio of 91%, growth remains strong. However, insurers face rising pressure to cut costs, enhance risk assessment, and maintain profitability. A key challenge lies in managing data.
On the Unstructured Unlocked podcast, Laura Drabik, Chief Evangelist at Guidewire, joined Tom Wilde, CEO of Indico Data, to discuss how artificial intelligence (AI) is reshaping underwriting by addressing inefficiencies tied to fragmented and unstructured information.
The Data Challenge in Underwriting
Underwriting relies on structured, high-quality data. Yet, much of the information carriers receive arrives in unstructured formats, such as emails, PDFs, and spreadsheets. Drabik noted that only 25% of broker submissions become written policies, and up to 60% are not reviewed at all — not due to lack of opportunity, but because underwriters are overwhelmed by fragmented data.
Submissions often land in centralized inboxes, creating bottlenecks. Extracting insights is time-consuming, while underwriters may also interact with as many as 15 different systems daily. This fragmented ecosystem reduces efficiency and leaves many submissions untouched.
The Role of Generative and Agentic AI
New technologies are helping address these challenges. Generative AI (GenAI) and agentic AI each bring distinct capabilities to underwriting.
-
Generative AI excels at summarization, context extraction, and decision support. It can:
- Triage broker submissions and flag promising opportunities.
- Enrich risk assessment by combining internal, external, and broker data.
-
Agentic AI acts autonomously to meet business goals. It can:
- Pre-clear submissions so underwriters see only viable risks.
- Orchestrate multiple data sources into structured frameworks.
- Recommend next steps, serving as a digital copilot.
Both Drabik and Wilde emphasized that human oversight remains essential. While GenAI handles interpretive tasks, agentic AI supports deterministic decision-making when guided by human supervision.
Strategies for Smarter Underwriting
The conversation identified several ways AI can improve efficiency:
- Focus on quality over quantity: Improving triage accuracy ensures underwriters work with complete, decision-ready submissions.
- Leverage external data: Only 40–50% of necessary risk data is broker-supplied. AI can pull in external datasets—from weather patterns to financial records—for more accurate models.
- Maintain human-in-the-loop systems: AI should support underwriters rather than replace them, with professionals overseeing and validating critical decisions.
Building Trust in AI
For AI to succeed in underwriting, trust is essential. Drabik and Wilde outlined three requirements:
- Establish ground truth: Define measurable benchmarks to evaluate AI effectiveness.
- Ensure transparency: AI systems must be auditable, with traceable data points and decisions.
- Start small and scale: Pilot programs in specific areas, like triage, before rolling out more broadly.
Smarter Technology for Smarter Decisions
The future of underwriting, according to Drabik, depends on making “smarter decisions, faster—with the right data at the right time.” With thoughtful implementation, generative and agentic AI can streamline workflows, improve risk modeling, and enhance decision-making.
Frequently Asked Questions
What risks exist if agentic AI is used without oversight?
It could make decisions based on incomplete or biased data, leading to compliance issues or inaccurate results.
How can underwriters build trust in AI beyond transparency?
Trust grows through consistent performance, peer adoption, training, and inclusion in the design and feedback process.
What integration challenges exist?
Legacy systems often pose hurdles. Insurers may face issues with data cleaning, workflow alignment, governance, and compatibility with modern AI tools.
Stay informed and ahead of the curve — explore more industry insights and program opportunities at ProgramBusiness.com.