Assist scaled from one-off modules to a customizable platform by redesigning the backend Builder UX. Institutions can now adapt knowledge bases to their own guidelines, and users can save preferences for faster workflows—reducing flagship maintenance costs by over 50%.
Role
Design Lead
Tags
B2B SaaS, Algorithm, AI Tool Builder
Timline
Q3 2025
Problem
As adoption grew, our flagship team could no longer satisfy every institution’s unique needs. They maintained separate versions of Assist modules for each client, which made updates slow, costly, and nearly impossible to scale. Needs varied not just across institutions, but also within teams and individual users—creating unsustainable complexity.
Process
We first stretched the existing system with small tweaks, focusing on low-cost solutions that reduced dev effort and onboarding friction. By partnering with the flagship team, we explored UX patterns to ease pain points without heavy builds and, in the process, identified which features were truly essential to plan.
(content board 아니면 미팅 캡처)
01
Institution / Customizable Knowledge Base
We modularized the GPT-embedded knowledge base so each institution could swap in their own guidelines during onboarding. Default values kept it optional, but flexibility made adoption easier and aligned Assist with client-specific standards.
02
User / Preferences
Using the familiar card UI from the Builder, we designed a lightweight preference-setting flow. Once saved, preferences applied automatically, sparing users from repetitive clicks while keeping the experience optional and intuitive.
03
Impact
For the flagship team, module development and maintenance costs were cut by more than half. Institutions gained Assist experiences that matched their guidelines, while end users benefited from customizable but streamlined workflows. This shift proved that we could satisfy diverse needs without bloating engineering scope, and client feedback consistently highlighted the improved flexibility.