Implementing AI Claims Triage for an Insurance Group

The Challenge

The organization ran on systems built a decade ago. Staff spent hours on manual data entry that should have taken minutes. Customer information lived in separate spreadsheets across three departments, and response times to client requests were getting worse, not better.

Our Approach

Research, Analysis, and Strategic Foundation

• Ran a market analysis covering target segments, competitor positioning, and where the real growth opportunities were
• Built customer personas from behavioral data, 20+ interviews, and third-party research
• Mapped the full customer journey and flagged the specific steps where people dropped off or got stuck

Design, Development, and Implementation

• Redesigned the key user flows based on the research findings, cutting three unnecessary steps from signup
• Built a testing framework so the team could ship changes weekly instead of monthly
• Added personalization based on what users actually did in the product, not just demographics
• Created dashboards that gave the leadership team a clear picture without requiring a data analyst to interpret them

Launch, Optimization, and Scale

• Scaled the campaigns that were working and shut down the ones that were not
• Expanded into two adjacent market segments that early data suggested would convert well
• Signed partnership deals with three complementary companies to reach new audiences
• Built retention programs focused on the first 90 days, where most churn happened

Key Takeaways

1. Growth worked best when marketing and product shared the same goals. Treating acquisition and retention as separate problems led to conflicting priorities.
2. Talking to customers every two weeks caught problems that the analytics dashboards missed entirely. Qualitative and quantitative data tell different parts of the story.
3. The teams that measured everything iterated twice as fast as the ones running on gut feeling. Setting up proper tracking before launching campaigns made all the difference.
4. Translating the website was not enough. Each market needed different pricing anchors, different trust signals, and sometimes a completely different value proposition.
5. Shortcuts in the codebase from year one cost us three months of rework in year two. Spending an extra week on architecture early would have saved far more time.

Explore Our Guides

Related guides from our knowledge hub: