Logility: Logility Expert Advisor (LEA)
Generative AI Chat Experience
Role
Product Designer
Researcher
Sector
B2B Saas, Gen AI
Supply Chain Platform
Tools
Mural, Photoshop
Figma, Figma Make
V0, Maze
Team
UX/UI Designer, Developer
Product Owner, Architect
Scrum Master, QA, Analyst
Introduction
As the lead UX designer, I was tasked with designing a natural-language interface for our enterprise supply chain platform, enabling users to ask high-level logistics questions and run AI-assisted analysis on real-time data. The company had an existing AI-based chat (DAI+) in one product, but we needed to update its look and feel, of the chat, improve usability, and expand functionality so it could work across multiple solutions. This new experience became Logility Expert Advisor (LEA).
-
Outdated look
-
Only source DAI+ data
-
Accessibility issues
-
Usability issues
-
Simple, modern design
-
Source data across all products
-
Ensure accessibility
-
Improve usability
Context & Constraints
This project came with unique challenges: we were in a post-acquisition transition, and I had no direct access to end users for research. To adapt, I relied on stakeholder interviews, analytics, and heuristic evaluations to identify pain points. I also conducted a competitive analysis of popular conversational AI tools to inform best practices. These steps allowed me to make informed design decisions despite limited user input.
Common Problems
Supply chain software is essential for companies to streamline and optimize their operations with users having to navigate multiple reports and dashboards to get answers. It can be a complicated, tedious and confusing process.
There are many common problems that users of supply chain software might face, prompting the implementation of an AI chat to assist:
Demand Forecasting
Issues
Shipping Delays
Inventory
Management
Supplier
Coordination
Cost Management
.png)
How can I accurately predict demand?
How can I avoid
and/or manage unexpected delays?
How can I keep
track of inventory levels and ensure timely restocking?
How can I manage relationships and communications with multiple suppliers?
What is the impact
of rising costs
in logistics, raw materials and labor?
Quality Control
Data Analysis
Customer
Service
Risk Management
Regulatory
Compliance
How can I ensure
consistent product quality across the supply chain?
How can I use data
to gain insights and make informed decisions?
How can I meet customer expectations for speed and quality
with changing demands?
How can I identify
and mitigate risks?
How can I navigate and comply with regulations across different regions?
UX Goals & Success Metrics
Implementing an AI chat can help to address many of these issues by providing real-time insights, automating responses, and offering predictive analytics to optimize supply chain operations. The core UX goal was to reduce cognitive load and make querying supply chain data as intuitive as messaging a teammate. Two main focus areas were identified:
Users should be able to:
-
Ask high-level logistic questions
-
Run AI-assisted analysis on their real-time data
-
Use natural language and familiar chat patterns
-
Find value in use for daily workflows (increase adoption)
Design should have:
-
Modern, intuitive, conversational interface
-
New features
-
Improved usability and accessibility
-
Ability to enable inquiries across multiple software solutions
-
Enhanced scalability for future products and features
Pre-Work & Process
Conducted stakeholder interviews, a heuristic evaluation, and competitive analysis. Gathered data and analyzed.
Redesigned the visual look of the chat using the design system. Use best practices and established practices.
Add new features, making sure to research purpose/functionality and what stage/release to implement.
Plan a testing process and feedback loop to gain insights for improvement and future releases.
Methods used since direct research wasn't possible:
-
Reviewed chat logs and analytics to identify common queries
-
Analyzed support tickets for pain points
-
Performed a heuristic evaluation of the existing chat
-
Conducted stakeholder interviews to understand business priorities
-
Conducted competitive analysis of other AI chats/tools
-
Built assumptions around user needs
-
Interviewed internal SME's to validate assumptions
Heuristic Evaluation
To get started, I analyzed the existing chat to understand how it worked and identify issues with the current design. I documented usability issues, including that the chat experience was limited in scope, was visually outdated, had accessibility issues, had poor visual hierarchy, suffered from inconsistent component usage and didn't support multi-product queries.
Previous Design: Issues
It is unclear what this small trash can icon with an "x" does. The design already has a close icon and as a persistent chat, it is not something a user would delete.
The "Write your message" prompt is hard to see due to the lack of contrast between the blue of the text on top of a darker blue background.
The dark blue Q & A is not centered in the chat window.
The color scheme of the chat is hard to view for long periods of time and the blue on blue is an accessibility issue.
The use of a transparent background here is distracting, making it harder for the user to concentrate on the relevant information.
The aqua outline buttons on a similar toned grey blue background is hard to see.
The dark send arrow is not visible on the dark background.
**Using a contrast checker, both the aqua arrows and the aqua outlined arrows fail with ratios of 4.14:1 and 1.71:1 respectively**
When you click on one of the questions above, the question/prompt comes into the message window instead of just going to the information.
The dark send arrow is replaced with an aqua arrow once a message is typed.
There is a close "x" icon here. The reason is unclear and it is hard to see (accessibility issue).
This color text (#FFFFFF) on the (#5391F7) colored message bubble fails the contrast checker with a ratio of 3.11:1.
Using multiple shades and tints of blue make it hard to decipher between message, message box and background.
Previous Design: Reaction Issues
Competitive Analysis
As part of the UX design process, I conducted a competitive analysis by evaluating multiple AI chat platforms to inform the redesign of our company's AI chat. I analyzed user flows, interface designs, conversational styles, and key features across leading AI chat tools to identify industry best practices and areas for improvement. This research provided valuable insights into what makes an AI chat feel intuitive, engaging, and trustworthy. By comparing strengths and weaknesses across different platforms, I was able to define key opportunities for differentiation and ensure our AI chat would deliver a seamless, user-centered experience that aligned with our brand's goals.
Key insights from other AI chat tools:
-
Clear conversational patterns (e.g., suggested prompts)
-
Persistent history for context
-
Visual feedback for system confidence
-
Error handling
-
Personalization features
User Flow
This user flow represents the MVP LEA chat representative of the first release. Over time this user flow will become more complex as additional features including data analysis, product filtering, Query ID, and credit management are added.
User Journey
This user journey shows the steps a user would take as they navigate the LEA Landing Page. The user would start by deciding a general product category, move to choose a specific product subcategory, and end when the user views the result from a specific question or command and chooses their next action.
-
Daily Standups with entire LEA Team
-
Teams Chat with Product Owner & Developer
-
Weekly meetings one-on-one with PO
-
As needed collaborations with Developer
-
Teams feedback chat with LEA Team & BC
Cross-Team and Additional Collaborations
New Chat Design (1st Release - Limited)
In the new design created in Figma, the chat feature is seamlessly integrated into the Logility platform, serving as a centralized hub for accessing the full suite of Logility products. The new design integrates the design system and establishes reusable patterns. LEA can be launched via an easily accessible icon. The home screen features product categories, suggested questions and prompts, a chat window that allows tagging specific products, and icons to minimize, maximize, or open in a new window. Additionally, a slide-out sidebar provides access to chat history. The home screen boasts a clean, simple, and user-friendly interface that adheres to accessibility guidelines.
Once the user selects a product category the chat will open with a prompt to then choose a subcategory to learn about. Choosing the subcategory will activate a list of relevant commands and questions.
The user has the option to minimize the chat which will then display the chat as an overlay contained to the righthand side of the screen:
Manage Chat History
Users needed the ability to manage their chat history by being able to edit the chat title, delete specific chats and clear the entire chat history. Being able to prioritize and order the chat was also implemented through the ability to star specific chats and drag to reorder.
Collapsed Tabs
LEA is heavily data driven with charts, tables and other diagrams. Users needed the ability to both expand a preview inside the chat window, and to open the full chart, table or diagram in a separate overlay.
Credit Management
LEA runs on a credit-based model so users would need the ability to view and manage the amount of credits they have in their account. I created an easy to access button located in the header that would trigger a dropdown that showed remaining credits, housed a credit usage log, gave the user the ability to set a credit cap and set a notification threshold. Credits can be color coded to indicate when a specific threshold had been met.
If the user activates the slider it triggers a dropdown where the user can enter specific information for the credit cap and/or the notification threshold. the user can then save the information which will then exit the dropdown. If the credits are above the set threshold it will display the number in primary blue. If the number is below the threshold it will display the number in a warning indicator orange.
If user chooses "View Credit Usage" or "Manage Credit Settings" they will be taken to an overlay that displays a usage log or editable credit settings.
Fork Chat
Fork chat enables the user to create a copy of the chat that includes all previous messages. The fork chat pop up is triggered by an icon underneath the chat response. Once the chat is forked, it will appear in the chat history and a fork icon will replace the chat icon.
Imbedded Feedback
In order to receive real-time ratings and feedback about the quality of the responses, I added thumbs up and thumbs down icons underneath each response. Each icon triggers an overlay form that prompts users to detail what they like or dislike, include a screenshot, learn more about what is collected and approve information shared.
Another Acquisition, Testing & Validation
Midway through the project, the company was acquired, and our development resources shifted, leaving us temporarily without front-end and QA support. Because of this, we released our in-progress version to a select group of clients and business consultants. I partnered closely with the consultants to gather early insights and used their feedback to begin shaping the next iteration while the team re-formed. Once the full team is restored and the next version is released, I will conduct structured usability testing to assess performance, uncover gaps, and drive the next phase of design improvements.
Impact, Next Steps and Future Release
Even with a limited early release, the feedback from business consultants provided actionable guidance. Their observations confirmed the value of the generative AI assistant in reducing manual steps and improving task initiation, while also revealing gaps in user guidance and conversational clarity. These learnings drove design refinements and informed the roadmap for the next phase. Broader usability testing is scheduled after the second release to measure effectiveness and uncover additional opportunities for optimization.
The first release demonstrated several meaningful improvements, including:
-
A scalable chat experience designed for multi-product integration
-
Reduced ambiguity in query handling through clearer UI patterns and conversation structure
-
A strong foundation for expanding LEA into additional AI-driven capabilities
-
Early validation from business consultants that informed the next iteration of the design
The updated design for the second release introduces a more intuitive and guided experience. It includes a redesigned home page featuring product-specific suggestions and recommended prompts, a new sidebar for selecting and switching between agents, and the ability for users to filter and favorite frequently used agents. The release also adds a dedicated preferences and settings area, along with a help and support page to improve user self-service.
This project demonstrated how generative AI can transform complex planning tasks into accessible, conversational interactions. It also pushed me to grow as a designer—particularly in creating clear, intuitive experiences around dense supply chain data. If given full research access, I would deepen validation through usability testing, task completion studies, and conversational flow analysis. The process ultimately taught me how to navigate constraints, extract value from limited data, and continue designing with intention even as circumstances shifted.
Success
Metrics
(Early Release)
Insight-Based
Foundation
Early consult validation
Scalable v1 chat experience delivered
Clearer query-handling patterns
Identified top usability issues
Defined plan for full usability testing
Usability
Direction
Roadmap
-
Received qualitative feedback from 10+ business consultants, informing the second release
-
Identified 3 critical friction points in the chat flow and addressed them in the updated design
-
Validated core workflows related to query handling, agent selection, and prompt guidance
-
Designed and delivered a scalable v1 chat experience that supported future AI-driven features
-
Developed a roadmap for full usability testing planned after the second release
