UI/UX Design · PM SalesForce
PMSF supports internal teams in managing product-related requests such as sampling, quotations, and approvals. Users consistently struggled with the Request Center, leading to high drop-off rates, incorrect submissions, and long onboarding times. I led user-centered design to simplify the workflow and align it with business goals.
Problem Overview
Users didn't know how to start or what information was required, especially for complex requests. The system was built for processes, not people.
Users abandoned submissions mid-process.
Operations team wasted time on clarification.
New users relied on colleagues instead of the system.
Goals & Success Metrics
Improve request accuracy and increase completion rate.
Reduce cognitive load; understand requirements without reading long documentation.
Research & Discovery
Stakeholder interviews with Product and Ops, 1:1 interviews with 8 internal sales users, shadowing real request submissions, and analysis of failed or rejected requests all pointed to the same root cause.
Users think in intent, not process: forms felt intimidating and "too technical."
Starting fresh was painful, so users duplicated past requests instead.
New users leaned on peers rather than trusting the system.
The Insight
How might we help users submit requests accurately without forcing them to understand the system logic? This led to an AI-assisted Request Flow that understands user intent, guides users step by step, and auto-fills and validates data dynamically.
Design Response
Instead of a list of request types, users see: "What would you like to request today?", with examples, or the option to upload a relevant document.
AI classifies intent into request type, required modules, and mandatory data, removing the need for users to decide upfront.
AI asks progressive questions only when needed: product category, quantity range, delivery urgency, reducing form fatigue.
The system builds the correct form, pre-fills known data, highlights required fields, and explains why information is needed.
Users review a clean, structured summary before submission, with the ability to re-edit, building confidence in their request.
Outcome & Impact
Submission errors reduced significantly.
New users ramped up faster on the system.
Higher trust in the system meant less manual clarification work.
Key Learnings
Good AI UX doesn't feel like AI. It feels like clarity. The most successful part of this project wasn't showcasing AI capabilities, but removing friction from the user's mental model. Shifting from "process-first" to "intent-first" design fundamentally changed how users interacted with the system, reducing cognitive load and building confidence.