InsureTech · Case study 05 of 12

Insurance Agent Portal

A major US insurer's agent portal had a 60-page training manual and still produced a 35% error rate on policy submissions. Agents spent more time navigating the tool than serving clients. The challenge was to reduce cognitive load so dramatically that the manual became unnecessary.

Insurance Agent Portal — InsureTech interface design by Yellowfirst
Client
Confidential Insurer
Year
2021
Industry
InsureTech
Role
UX Research · Interaction Design · Learnability

The challenge

The portal had been built screen-by-screen over 12 years by different vendors. There was no information architecture — screens were ordered by database table, not by agent workflow. Agents needed to hold 6-8 data points in working memory simultaneously while switching between tabs. New agents took an average of 4 months to become proficient.

The solution

We restructured the entire portal around the agent's actual job: quoting, binding, and servicing policies. A persistent context rail keeps all relevant client and policy data visible regardless of which step the agent is on. Smart defaults and relevance filtering eliminate 60% of the fields agents previously had to evaluate on every submission. In-line guidance replaces the training manual with contextual help surfaced at the moment of need.

Our process

  1. 01 Cognitive Load Assessment

    Measured working memory demands at each screen using NASA-TLX workload scale. Identified 4 screens with unacceptable cognitive cost.

  2. 02 IA Restructure

    Re-sorted 85 screens into 4 primary workflows based on agent task frequency and dependency analysis.

  3. 03 Context Rail Design

    Designed a persistent client context panel — always-visible policy summary that follows the agent through every step.

  4. 04 Smart Default System

    Built a relevance engine that pre-selects form fields based on policy type, agent history, and client profile. Reduced visible fields by 60%.

  5. 05 Learnability Testing

    Tested with 12 new agents who had never seen the portal. Measured time-to-competency against baseline. Iterated twice.

Outcomes

  • 35%→4%Error rate on policy submissions
  • 4 mo→3 wkNew agent proficiency timeline
  • 60%Reduction in visible form fields
  • 0 pagesTraining manual required post-launch

Technology

  • React
  • REST APIs
  • Azure AD

Read more