Underwriting in action: real-world Agentic AI use cases 

Send recently hosted a webinar in collaboration with InsTech titled, ‘Underwriting in action: Real-world Agentic AI use cases’, during which our Head of Revenue for North America, Brad Tabor, joined a panel of experts – Carey Geaglone, Senior Principal Insurance Practice at Datos Insight, and Amy Nelsen, Head of UW Operations for US Middle Market at Zurich North America – to share their insights on how AI is reshaping underwriting.  

During the session, Brad covered the evolution of AI from basic extraction to advanced, agent-driven intelligence, removing the ‘Google’ step, and the importance of using AI while keeping the underwriter firmly in the loop. 

Giving the audience a practical, real-world view of the current applications of agentic AI, Brad raised three key use cases.  

Intelligent submission triage at scale

The first common use is deploying AI to scan entire submissions for key details and flag important information, such as historic claims. Underwriters can work more efficiently, no longer combing through pages and pages of information, loss runs etc, and getting to the facts, quicker.  

Using AI to provide better context and a better understanding of the risk

The second was removing what he called ‘the Googling side of underwriting’ by using agents to search public information, recent news and other data banks to help underwriters understand the scope of risk in less time.  

This manual task can take hours, without certainty that the risk is within appetite, but with agentic AI, quote turnaround times have vastly improved, which could therefore increase the submission-to-quote ratio. 

Reclaiming underwriter time to focus on high-value work

Finally, AI can significantly reduce manual work, such as the time spent building an underwriting file. This documentation is vital for explaining the risk to chief underwriting officers and senior underwriters, yet it is slow and repetitive to produce. As Brad noted, ‘Underwriters don’t want to spend time documenting their file, they want to spend their time crafting the deal, talking to brokers.’ With AI, much of this process can now be automated, ensuring the file is complete and accurate for senior staff while freeing up underwriters to focus on higher-value work. 

When asked about designing these systems responsibly, Brad emphasised that AI must remain easy for underwriters to understand, with firm guardrails to ensure it stays aligned with underwriting judgment. For example, clear prompts reduce the risk of hallucination, and companies should carefully review outputs while the technology continues to mature.

He also noted that AI is only valuable if it’s auditable – AI that can’t be explained is AI that can’t be adopted.

Brad concluded “make sure you are putting a human touch around your work, using AI in a way that is going to truly impact your business without taking the underwriter out of underwriting”, reminding the audience that technology cannot replace the underwriter, and that the human touch remains essential. 

Catch the full panel discussion with insights from all speakers in the link below:

Underwriting in action: Real-world Agentic AI use cases