Duck Creek Case Study

Goal

The collaboration focused on the implementation of an AI-assisted predictive model creator used by actuaries to calculate insurance risks. Our main goals were to integrate these new AI-assisted features seamlessly into the existing Duck Creek Machine Learning software, and create a user-friendly, low-code environment for all available features.

Challenge

When we joined the project, the AI risk modeling tool relied on a lot of manual coding. We created a more intuitive design that required less technical know-how from the user end. Our adaptability, deep dive into AI, and holistic approach to interconnected product modules led to a smooth and quick collaboration.

Outcome

By refining the UX and delivering key research, we transformed Duck Creek’s ML platform into an accessible tool—paving the way for a brand-new business segment.

1 /

Discovery

How we started

First, we focused on understanding the product's broader context. It included user needs, their tasks, mapping the environment, and exploring technical opportunities.

Stakeholder interviews, market research, and brand direction definition were also an integral part of this phase. These findings helped us craft a fresh identity for the newcomer product within the DuckCreek family.

Ideation session

A digital whiteboard showing a colorful flowchart titled "Exploring the existing information architecture," with labeled sections for workflow, steps, data transformations, data view, and algorithms.

2 /

New feature

From sketch to successful solutions

The “generic” feature was entirely new to the product, requiring a ground-up approach. Due to this, the design and development took 6 weeks. This also included ideation sessions and sketching, concept testing, and the delivery of the final screens.

We built a robust infrastructure supporting Duck Creek's Machine Learning platform’s experiments. Taking inspiration from Google Drive's file system, our design allowed for efficient file organization, navigation, search, and permissions management.

Simultaneously, we prioritized integrating it with the “Canvas” feature to harmonize functionalities and enhance user experience in the overlapping areas.

This phase included

1. Moodboards
2. Look & Feel
3. Concepts
4. Information Architecture
5. Hi-fi Wireframe

Chosen moodboard

A moodboard showing the font and colour palette selected and some additional images to have a better impression of the overall mood intended

3 /

Canvas Design

Perfecting the predictive model canvas

In the final 10-week phase, we focused on the “Canvas,” the central area for predictive model creation in Duck Creek's Machine Learning platform. The “Canvas” feature was important because it allowed the users to view the experiment they created in the “Generic” view.

This interactive component required thorough testing and multiple iterations. We used clickable prototypes to verify functionality and get fresh feedback from stakeholders and potential users. Ultimately, we made this interaction-heavy part to offer clear model overviews, fast execution, and customizable settings.

This phase included‍

1. Concepts
2. Artboards (screen designs)
3. DocumentationUI kit

4 /

The Outcome

Smart tech, strategic UX

We achieved significant milestones during the project, including mapping the product structure, establishing brand guidelines, and creating design assets. From the start, we gathered a lot of insights, which we used to create a feasible timeline and prioritize features.

Through strategic design and thoughtful UX improvements, we turned a complex, code-heavy product into a seamless, user-friendly platform—helping Duck Creek bring cutting-edge AI closer to its customers.

Deliverables‍

Design artifacts.
Generic part design.
The Canvas design.
UI kit with all assets for future expansions.

Documentation

Insights gathered during collaboration.
Evidence supporting design decisions.