Ask ARIES: A digital assistant, powered by AI, lets users ask HR questions, find documents, explore laws and regulations and more.
Company
Mineral, Inc (now Mitratech)
Role
Product Design Lead
Introduction
Project Overview
ARIES is an AI-driven chat assistant integrated into the Mineral Platform. It leverages natural language processing to answer Tier 1 (low-complexity) HR compliance questions. Responses provide summarized insights and direct users to relevant tools and resources within the platform.
Role
As the Product Design Lead, I spearheaded the UI design from inception to completion. I translated initial business requirements into design solutions, prototyped and tested concepts, and delivered high-fidelity designs for production.
Objective
Mineral aimed to establish itself as a leader in predictive and personalized HR solutions by harnessing AI technology.
Problem Statement
User Pain Points
A dependency on office hours: The HR Assistant provides 24/7 access to HR advice.
A "submit and wait" ticket system: Delivering faster, more personalized answers compared to ticket submissions or manual content searches.
Sifting through a massive amount of content: Leveraging Mineral’s extensive HR knowledge base by training an LLM.
Product Design Challenges
Determining how to integrate the AI assistant within an already feature-rich platform.
Designing for seamless integration without disrupting existing workflows.
Balancing innovation with technical constraints and MVP scope.
Addressing limitations regarding AI model access to platform data and content.
Maintaining trust and human oversight in sensitive HR environments.
Discovery Phase
Research
I conducted thorough research on best practices for chat-based AI interfaces across various industries. I actively engaged with early versions of ChatGPT, exploring the evolving UI and its user experience. I documented my findings in a shared Figma board, curating design inspirations for team discussions.
Ideation
In the sketching phase, I explored diverse launch and interaction models for the assistant, from minimal UI triggers to comprehensive dashboard integrations. This stage encouraged creativity and out-of-the-box thinking, laying the groundwork for initial concepts.
Sketches: They may not be for everyone, but they help me think through possibilities. I often snap a photo of my sketches and paste them directly in Figma next to my design. They are great way to reference core ideas throughout the project.
•••
Considerations
Given the evolving nature of AI capabilities, the design needed to remain adaptable to technical limitations. I navigated overlapping platform features, such as an existing help chat icon, prompting multiple stakeholder reviews to ensure cohesive integration.
User Testing
Testing Methods
Testing involved flat wireframes and clickable prototypes, conducted via Zoom with internal users and clients. This iterative process validated design decisions and aligned the assistant’s functionality with real-world needs.
Insights Gained
User feedback influenced key design elements, including widget placement, layout structures, and feedback flows. These insights refined the assistant’s UX to better meet user expectations.
Above: Three options for the placement of the Chat Icon. Launching it from the bottom right was ultimately not in scope mid-way through the project. We reviewed three placements with users. Option A was chosen, which was a dedicated chat icon within the utility navigation.
•••
Design and Build
Use Case Alignment
Once the core design and interaction model were established, I transitioned to detailed design work. Jira stories, created by the Product Manager, outlined specific use cases and requirements. I annotated Figma screens to ensure seamless handoff, visually mapping flows from start to finish.
I developed desktop and mobile designs for over 60 stories, ensuring each design was clear enough for PMs and engineers to implement with minimal oversight. I remained engaged throughout development, attending demos, QAing the product, and filing bug reports to uphold design integrity.
Team Collaboration
The project team included a Business Analyst, Lead Product Manager, front- and back-end engineers, HR specialists, AI content curators, and senior marketing leads and myself. Each stakeholder played a critical role in refining the assistant’s functionality and ensuring alignment with business goals.
Above: Sprint demo (recording) lead by engineering team. As a designer, it's really satisfying to see the UI take shape on the engineering side. While I was there to field questions on the user experience and design side, I often heard great feedback from stakeholders.
•••
Final Deliverables and Outcomes
Deliverables
✔️ High-fidelity designs for all user stories and interaction flows.
✔️ Design system updates and documentation within Quartz Design System.
✔️ QA and engineering collaboration to ensure product accuracy.
Lessons Learned
This project highlighted the value of over-communicating design decisions to reinforce trust and collaboration across teams. Although not all concepts shipped with the MVP, designing for long-term scalability allowed the team to future-proof the product.
Working on ARIES deepened my passion for AI-driven design and marked a significant milestone in my career. It underscored the excitement of contributing to innovative technology and reaffirmed my interest in designing for AI-based solutions.
Results
The assistant has since successfully launched, enhancing the platform's accessibility and user engagement.
The above link includes a product overview and promotional video created by Mitratech, the company that acquired Mineral, Inc. in early 2024.
•••
Screenshots
Above: This flow showcases the interaction when a user rates a response given by the HR Assistant. A user could chose rate the quality of the response to leave feedback or not. This data created an internal customer support ticket, and supported our HITL (Human in the Loop) initiative.
•••
Above: Employee count is important when determining what federal and state laws appy, especially small businesses. This flow shows a UI interaction to update company information so that the HR Assistant could provide the most accurate responses. We ultimately designed and placed this piece of UI in multiple areas of the platform.
•••
Above: Desktop screen display of inline chat on a dedicated chat page. These screens defining how a question is displayed, how a response/summary is displayed, how externally linked documents and action items are displayed.
•••
Project Summary of Responsibilities
Discovery
Understand Business Requirements (BRDs), timeline, scope and KPIs.
Research trends and best practices of advanced chat and AI technologies.
Planning
Establish a design timeline, communicate design deliverables, review with stakeholders.
Work with Project Manager to establish user stories and use cases.
Design & Prototyping
Sketches and drawings of various concepts.
Wireframes to review and test and learn with users.
High fidelity designs of large concepts; review and test and learn with users.
Share and review with stakeholders.
Navigate feedback and pivots to identify MVP.
Final Deliverables
Create high-fidelity designs for all flows, interactions for user stories in Figma.
Document UI and incorporate new design elements within the Quartz Design System.
Demo interactions with engineers; collaborate and QA the build.