The Case Study
Discover how to design an AI-driven claims triage system for a regional insurer using Claude. Learn to handle intake by classifying claims, assessing severity, and routing accurately within regulatory AI governance frameworks. This lesson helps you understand stakeholder perspectives, compliance requirements, and the scope constraints in a real insurance scenario.
Insurance is a heavily regulated industry where AI is spreading quickly, especially in claims. This case places us as a Forward Deployed Engineer embedded with a regional insurer that wants Claude to help its claims team handle incoming claims faster.
This lesson presents the case as a customer brief. It describes the insurer, how claims are handled today, what the insurer is asking for, the governance rules it operates under, and what the engagement covers.
The case begins where every claim begins, in the intake team’s inbox.
The insurer and its claims team
Harborline Mutual Insurance is a regional insurer that sells auto and homeowners insurance in six states. It covers about 400,000 policyholders and runs three claims centers. The largest, in Columbus, Ohio, handles about 3,000 new claims in a typical week.
When something goes wrong, such as a car accident or a burst pipe, a policyholder reports it to the insurer. That first report is called the first notice of loss (FNOL). At Harborline, most first notices arrive by email, often with attachments such as photos, police reports, repair estimates, or a completed claim form saved as a PDF.
A team of 22 intake specialists in Columbus reads each new claim and sends it to the right place. Claims are then worked by adjusters, the insurance professionals who investigate each claim, determine what the policy covers, and decide what the insurer pays. Each adjuster ...