Intelligent Warehouse Process Automation (IWPA)
.Client
Alibaba Group
.My Role
UX Design, Product Design.
.Our Users
Support Associate, Order Picker, Inventory Control Specialist.
.Timeline
06.2024-09.2024, Launched 11.2024
.Team
YueYang - PM
BaoWang - Product Leader
QianNing - Engineer
ShiLang- Engineer
Rengqi Han- Data Scientist
.Overview
Warehouse operations were inefficient, with field efficiency constrained by outdated workflows on the digital platform. To address this, Alibaba sought an AI-driven solution leveraging digital interactions to boost productivity. This smart warehouse platform empowers workers, enhances efficiency, and lowers warehouse costs.
Background
Baseline Metrics
We conducted system data analysis and surveys across 40+ warehouses nationwide, gathering key performance metrics.
Problem statement
The workflow is highly interdependent. An error in one step can disrupt the entire process. Operations rely on traditional manual methods, lacking automation and intelligent optimization.
Deisgn goal
Using AI
AI holds great potential; we aim to fully leverage our existing capabilities by FY 2025
Deisgn goal
Reduce operation time and shorten the ready-to-ship timeframe.
Project timeline
The AI project is the core focus. In the first two months, we developed and iterated on new features. In the third month, we explored new functionalities to define potential future directions.
Field research
Interview & roleplay
Beyond desktop research, field operations require a user-centered approach. To deeply understand the scenarios , I conducted in-depth interviews and facilitated task-driven, immersive roleplays to observe the real operational habits of three key stakeholders.
This process helped formulate hypotheses about core problems and insights. We identified pain points in the operational workflow that contributed to increased time and error rates, some of which could be mitigated through interface iteration.
Observation
Observing real workers in action, we found high communication costs due to the warehouse’s size and frequent staff turnover. Repetitive questions wasted time for both workers and responders, creating inefficiencies.
Affinity map
We organized the interview notes and used affinity mapping to categorize key insights into major dimensions such as interface, communication, and operations. We analyzed influencing factors and the number of affected users.
Retrospective
I faciliated retrospectives with PMs and engineers, bringing my insights to the discussion. Based on our collective insights, we brainstormed and ultimately decided to focus on two key challenges for new features, particularly in optimizing picking time and exception handling.
Define prompts
Align team visions and narrow down key challenges to ensure a strategic approach to new feature development.
Cross-team brainstorms
Facilitate brainstorming sessions, aiming to gather insights from diverse perspectives and drive innovative solutions.
Led design plan doc
Create and structure design plans, ensuring alignment on priorities, clear direction for development.
Insight
Problem solving
I consulted with engineers and learned that, at Alibaba, shelf location and mobile GPS data are already integrated into the system using IoT technology. Given the availability of this data and AI’s potential for real-time processing, I believe we can leverage it to enhance location-based guidance.
Based on field observations and roleplays, we identified two main factors that slow down the picking process— visual hotpot and habitual focus zone.
User flow
I explored multiple approach ideas, created user flow sketches, and discussed them with the UX team and my mentor. Through our discussions, we determined that this workflow is both user-centered and technically practical.
Iteration
Before - Item Picking
Cons:
01 / Displays all information without providing helpful hints.
02 / Picking the wrong item often occurs.
After - Item Picking
Solution:
01 / Calculates the optimal picking order and routes in real time to minimize travel distance.
02 / Provides visual hints for the item's relative location on the shelf.
Pros:
01 / AI optimizes routes while keeping pickers in control of the picking order.
02 / Provides visual cues for routes and item locations.
03 / Provides both a 2D map and a real-time navigation map
04 / Integrates new flow to prevent picking the wrong item.
Flow 1
I improved efficiency through two approaches: optimizing shelf organization and enhancing item location guidance. In this iteration, I also left an entry point for integrating new features my colleague is developing.
Insight
Solutions
I consulted with engineers and learned that, at Alibaba, shelf location and mobile GPS data are already integrated into the system using IoT technology. Given the availability of this data and AI’s potential for real-time processing, I believe we can leverage it to enhance location-based guidance.
User testing
To ensure usability, we conducted user testing and interviews in the warehouse to validate and inform the selection of the design approach moving forward.
Since field operations require minimal distractions, voice input is a good option. We conducted tests to assess whether voice commands would cause interference when multiple workers were using them in the aisles. The results showed that voice input was effective in the aisle environment, and workers responded positively to its use.
We also discovered that due to workers' backgrounds, they have strong accents, which standard voice models struggle to accurately recognize.
User flow
I explored multiple approach ideas, created user flow sketches, and discussed them with the UX team and my mentor. Through our discussions, we determined that this workflow is both user-centered and technically practical.
Iteration
Before - Ask and do action
Cons:
01 / Wastes time for both the asker and responder.
02 / Long walks between workers, the field, and PCs.
03 / High onboarding turnover leads to repeated basic questions.
After - Ask and do action
Solution:
01 / Implement voice input for faster interactions.
02 / Utilizes GenAI for real-time solutions.
Pros:
01 / Easy to use while walking in the field.
02 / Provides real-time responses.
03 / Enables quick actions in field operations.
04 / Designed AI cards with enhanced capabilities.
Flow 2
Through two approaches to improve efficiency, one is optimizing the way the shelf, and second is helping locate items in the shelf. In this iteration I also leave the entry to input new features which my college is working on.
Insight 3 - How can we improve PC operation efficiency?
Since multiple windows overload the screen, users often misclick and struggle to find the page they need.
Iteration
Solution 3 - Introduce AI filter and mutilpanel view
I redesigned the interface to integrate a new AI-powered filtering feature.
Before - Exception forms
Cons:
01 / Filters are redundant.
After - Exception forms
Solution:
01 / AI-powered search bar.
02 / Split-view panels for better navigation.
Pros:
01 / Reduces search steps.
02 / Improves information hierarchy and clarity.
Flow 3
Through two approaches to improve efficiency, one is optimizing the way the shelf, and second is helping locate items in the shelf. In this iteration I also leave the entry to input new features which my college is working on.
Integration
To enhance ChatAI's capability as a versatile internal tool for various business units, I addressed integration challenges across teams.
I collaborated with technical teams to develop a universal design pattern but faced issues like data security concerns and siloed management between overseas and local teams.
To resolve this, I designed three card solutions: a comprehensive option requiring broader data permissions and collaboration, and a basic option for teams preferring minimal resource and data involvement.
The three bundles offer flexibility tailored to team preferences and resource commitments.
Impacts
