lan hoang

Keeyu

Research, platform redesign, design system, agent experience2026Credits
Refract
Visit website: refract.group (opens in a new tab)

Keeyu helps large retailers deal with the messy work that starts after an order is placed.

Overview

Problem
Keeyu was already good at spotting what was going wrong with an order. The next step was fixing it in the same place: many customers still exported their data to spreadsheets to finish the work, where Keeyu couldn't see it or help automate it.
What I did
At Refract, I designed and ran the research, restructured the platform around the work each team does, designed Keeyu's AI workflows and Agent experience, and built the design system the product now builds on.
Outcome
The designs were built with Keeyu's engineers. I tested them again with the same customers after development, and kept iterating from their feedback.

Selected work

Order Tracker

Shows the chain of events behind an order's status, so nobody has to interpret a single status field.

Workflow Manager

Lists every step the Agent took on an order, so a customer can check exactly what it did.

Agent actions

The Agent stops and asks when a step needs a person, and a choice can be undone.

AI composer

A workflow emails the customer about a problem before they notice it, and a person can edit the draft.

Onboarding guidance

Help appears inside the setup task, pointing to the field or step that is missing.

Design system

Defines type, colour, components and patterns once, so new screens inherit them.

Password required

Enter the password to read the full case study.

Don't have it?