Simplifying thousands of legislated classifications into an AI tool.
Each WorkCover insurance applicant is assigned one of 9,907 business classifications manually, affecting their annual payment. Misclassifications can cost a business significantly over time.
The preferred state has always been for self-classification, but how do you manage all that complexity?
We used the structure of the legislation to influence the nagivation and other UI design.
Applying service design principles to gathering requirements was critical on this project. The classification process is legislated, so we had to use it as a guide.
Understanding the overall steps in a journey format helped us understand how to transform the legislation into a self-service digital experience.
Unintuitive definitions were addressed by using metadata and AI to give users multiple chances to recheck their classification.
Putting the law at the heart of our decision making worked, with an AI-supported prototype with 90+% accuracy.
Straight-through-processing was finally proved to be a feasible option, and my work was showcased to multiple industry partners within the WorkCover industry.
This was a challenging and fun project, with a deep level of subject matter complexity and an opportunity to influence the entire solution from the UI down to the algorithms themselves.