Complete redesign with new user flows to improve feature discoverability and creation of design system for scalability.
Complete redesign with new user flows to improve feature discoverability and creation of design system for scalability.
Timeline
Timeline
Sep - Dec 2025
Sep - Dec 2025
Sep - Dec 2025
Team Members
Team Members
Robert Bowser (Product Designer), Tanvi Sanandiya (UX researcher), Shreyash Hamal, Aria Han, Myat Paing (Cofounder and Developer)
Robert Bowser (Product Designer), Tanvi Sanandiya (UX researcher), Shreyash Hamal, Aria Han, Myat Paing (Cofounder and Developer)
Robert Bowser (Product Designer), Tanvi Sanandiya (UX researcher), Shreyash Hamal, Aria Han, Myat Paing (Cofounder and Developer)
My Role
My Role
Product Designer
Product Designer
Product Designer
PersistOS is a privacy-first startup that focuses on contextual intelligence. HeyContext is an Autonomous Work Platform that stems from PersistOS’ core technology and is at an early stage of product development.
The Challenge:
Market Repositioning: With a recent pivot in targeted use cases, the redesign needs to accommodate enterprise users and relevant expectations while preserving the user-friendliness of consumer-facing digital products.
Usability Optimization: Previous user testing and heuristic evaluation indicated opportunities to enhance overall usability to better address user needs.
Feature Discoverability: Redesign should surface key functionalities which were not discoverable or accessible to users.
Limited Resources: With only 2 designers onboard working part-time, we had to allocate our efforts strategically.
We Delivered:
Design Foundation: Developed a comprehensive design system along with a refreshed brand identity, and established guiding design principles to ensure intentional, values-driven design decisions and support future product scaling.
UX Restructuring: Reimagined user story and platform architecture; new user flow illustrated with high fidelity wireframes and Figma Make prototypes.
PersistOS is a privacy-first startup that focuses on contextual intelligence. HeyContext is an Autonomous Work Platform that stems from PersistOS’ core technology and is at an early stage of product development.
The Challenge:
Market Repositioning: With a recent pivot in targeted use cases, the redesign needs to accommodate enterprise users and relevant expectations while preserving the user-friendliness of consumer-facing digital products.
Usability Optimization: Previous user testing and heuristic evaluation indicated opportunities to enhance overall usability to better address user needs.
Feature Discoverability: Redesign should surface key functionalities which were not discoverable or accessible to users.
Limited Resources: With only 2 designers onboard working part-time, we had to allocate our efforts strategically.
We Delivered:
Design Foundation: Developed a comprehensive design system along with a refreshed brand identity, and established guiding design principles to ensure intentional, values-driven design decisions and support future product scaling.
UX Restructuring: Reimagined user story and platform architecture; new user flow illustrated with high fidelity wireframes and Figma Make prototypes.
PersistOS is a privacy-first startup that focuses on contextual intelligence. HeyContext is an Autonomous Work Platform that stems from PersistOS’ core technology and is at an early stage of product development.
The Challenge:
Market Repositioning: With a recent pivot in targeted use cases, the redesign needs to accommodate enterprise users and relevant expectations while preserving the user-friendliness of consumer-facing digital products.
Usability Optimization: Previous user testing and heuristic evaluation indicated opportunities to enhance overall usability to better address user needs.
Feature Discoverability: Redesign should surface key functionalities which were not discoverable or accessible to users.
Limited Resources: With only 2 designers onboard working part-time, we had to allocate our efforts strategically.
We Delivered:
Design Foundation: Developed a comprehensive design system along with a refreshed brand identity, and established guiding design principles to ensure intentional, values-driven design decisions and support future product scaling.
UX Restructuring: Reimagined user story and platform architecture; new user flow illustrated with high fidelity wireframes and Figma Make prototypes.
Design Questions
How might we surface and translate complex AI generation capabilities into accessible and effective features serving users of varying technical expertise and professional backgrounds?
How might we integrate discovery, design, and delivery phrases in an agile startup environment while maintaining design quality and development momentum?
Research
My work at HeyContext began by synthesizing insights from 9 baseline interviews with individuals of varying AI use that asked about their existing frustrations, expectations, and wishes for AI platforms.
Key Insights:
Repetitive and lengthy prompts create user friction. ("Writing long prompts again and again to provide context is frustrating")
Users desire proactive, anticipatory assistance. (“AI acting as a proactive assistant rather than a reactive tool would be a great improvement”)
Users would like it embedded in the actual workflow rather than something they have to go to.
From this, we also identified the following mental models:
Ambient AI Agents: users expect AI agents to be autonomous, event-driven, and persistently context-aware such that no manual prompting is needed.
Outcome-focused: users’ evaluation of AI platforms is largely dependent on generation quality. Ideally, generated deliverables should match both the immediate user input and the larger context, uphold high accuracy and quality, while also allowing for rapid iteration.
Research
My work at HeyContext began by synthesizing insights from 9 baseline interviews with individuals of varying AI use that asked about their existing frustrations, expectations, and wishes for AI platforms.
Key Insights:
Repetitive and lengthy prompts create user friction. ("Writing long prompts again and again to provide context is frustrating")
Users desire proactive, anticipatory assistance. (“AI acting as a proactive assistant rather than a reactive tool would be a great improvement”)
Users would like it embedded in the actual workflow rather than something they have to go to.
From this, we also identified the following mental models:
Ambient AI Agents: users expect AI agents to be autonomous, event-driven, and persistently context-aware such that no manual prompting is needed.
Outcome-focused: users’ evaluation of AI platforms is largely dependent on generation quality. Ideally, generated deliverables should match both the immediate user input and the larger context, uphold high accuracy and quality, while also allowing for rapid iteration.
Through desk and market research, we also discovered the critical gap: no existing AI agentic product combines autonomous multi-agent coordination with persistent organization memory, structured deliverable generation, all the while without manual prompts.
Through desk and market research, we also discovered the critical gap: no existing AI agentic product combines autonomous multi-agent coordination with persistent organization memory, structured deliverable generation, all the while without manual prompts.
Early platform development focused on rapid feature delivery, with the interface generated through AI-assisted development tools. To evaluate the existing experience, we conducted 7 moderated usability testing sessions and a comprehensive UI audit.
While critical usability issues were immediately addressed with the development team, further analysis revealed the following critical pain points:
poor feature discoverability
inconsistent interaction patterns
lack of visual identity.
Early platform development focused on rapid feature delivery, with the interface generated through AI-assisted development tools. To evaluate the existing experience, we conducted 7 moderated usability testing sessions and a comprehensive UI audit.
While critical usability issues were immediately addressed with the development team, further analysis revealed the following critical pain points:
poor feature discoverability
inconsistent interaction patterns
lack of visual identity.
Synthesis
Based on user research, competitive analysis, and company mission, we established 3 core values to guide design decisions and product development.
Synthesis
Based on user research, competitive analysis, and company mission, we established 3 core values to guide design decisions and product development.
Design System
With this in mind, we introduced a new design system intended as a critical framework to accelerate development, align cross-functional teams, and enable incremental, systemic implementation of the redesign.
Scalable Component Library (140+ Elements): Established an extensible system of reusable components—spanning basic UI elements, iconography, and design patterns that empowers designers and developers to build with better efficiency.
50+ Style Specifications: Systematized typographic hierarchy, gradients, and semantic color usage, ensuring accessible and cohesive visual experiences.
Variable Collections & Design Tokens: Token-based architecture with native light/dark modes, establishing a single source of truth for design decisions and ensuring design-development consistency.
Parallel to UI kit and design pattern creation, we also produced high-fidelity prototypes for the two core user journeys: artifact creation via conversation and context-preserved editing upon return.
Parallel to UI kit and design pattern creation, we also produced high-fidelity prototypes for the two core user journeys: artifact creation via conversation and context-preserved editing upon return.
Design Highlights
Design Highlights
Handoff & Implementation
Delivered Figma file including Design System UI kit and high fidelity prototypes, established a phased rollout strategy facilitated by Figma MCP.
Exploring Figma Make to optimize implementation efficiency and ensure accurate translation between design specifications and code implementation.
Next Steps
Expand the component library and design patterns while developing documentation that supports scalable product development.
Obtain data on user feedback and emergent needs as ways to evaluate design system and iterate accordingly.
Handoff & Implementation
Delivered Figma file including Design System UI kit and high fidelity prototypes, established a phased rollout strategy facilitated by Figma MCP.
Exploring Figma Make to optimize implementation efficiency and ensure accurate translation between design specifications and code implementation.
Next Steps
Expand the component library and design patterns while developing documentation that supports scalable product development.
Obtain data on user feedback and emergent needs as ways to evaluate design system and iterate accordingly.
Evaluation
Redesign Success is expected to be quantified through following metrics:
Adoption metrics: Component usage frequency in design and production, feedback from designers.
Efficiency indicators: Design-to-development cycle time and direct feedback from developers.
User Satisfaction: Prospective user feedback collection.
Assessment methods include analytics tracking, regular cross-functional review sessions, and moderated usability testing.
Strengths and Limitations
The redesign is a through-through solution that incorporates company values and addresses identified user needs. It also establishes the foundation for systemic design and will continue to evolve as the company grows.
Limitations to our redesign is the absence of user validation of implemented features, which we hope to obtain via user testing in the near future. Personally, a hurdle to my learning is the lack of mentorship from experienced designers.
Feedback Status
Internal feedback (developers, stakeholders) is positive regarding workflow improvements and visual identity. External validation with target enterprise users is scheduled for upcoming sprints, which will provide critical assessment of user-facing design effectiveness.
Key Learnings:
🌟 Implementation Ownership: For this internship, I anticipated learning about handing off and maintaining the design system as an in-house designer. In this case, as our developers are not specialized in frontend, I decided to take on implementation responsibility to maintain design fidelity. This also reflects industry trends where UX designers increasingly own execution alongside strategy.
🌟 Trial and Error in AI Interactive Patterns: Analyzing recently launched platforms like Microsoft Copilot and Gemini Enterprise revealed that established AI interface patterns aren't always optimal as they are also at early stages of product development. Our user research-driven, iterative approach produced solutions that competed favorably with industry giants.
🌟 Professional Presence & Mentorship: While the startup's collaborative culture valued all contributions, establishing credibility through cross-functional engagement and relationship-building strengthened design impact.
🌟 Proactive Communication: Working in a small-team, self-directed capacity required intentional communication—sharing progress, requesting updates, and asking questions to uncover collaboration opportunities. Overcoming initial hesitation with direct inquiry strengthened team alignment and revealed colleagues' openness to transparent collaboration.
Evaluation
Redesign Success is expected to be quantified through following metrics:
Adoption metrics: Component usage frequency in design and production, feedback from designers.
Efficiency indicators: Design-to-development cycle time and direct feedback from developers.
User Satisfaction: Prospective user feedback collection.
Assessment methods include analytics tracking, regular cross-functional review sessions, and moderated usability testing.
Strengths and Limitations
The redesign is a through-through solution that incorporates company values and addresses identified user needs. It also establishes the foundation for systemic design and will continue to evolve as the company grows.
Limitations to our redesign is the absence of user validation of implemented features, which we hope to obtain via user testing in the near future. Personally, a hurdle to my learning is the lack of mentorship from experienced designers.
Feedback Status
Internal feedback (developers, stakeholders) is positive regarding workflow improvements and visual identity. External validation with target enterprise users is scheduled for upcoming sprints, which will provide critical assessment of user-facing design effectiveness.
Key Learnings:
🌟 Implementation Ownership: For this internship, I anticipated learning about handing off and maintaining the design system as an in-house designer. In this case, as our developers are not specialized in frontend, I decided to take on implementation responsibility to maintain design fidelity. This also reflects industry trends where UX designers increasingly own execution alongside strategy.
🌟 Trial and Error in AI Interactive Patterns: Analyzing recently launched platforms like Microsoft Copilot and Gemini Enterprise revealed that established AI interface patterns aren't always optimal as they are also at early stages of product development. Our user research-driven, iterative approach produced solutions that competed favorably with industry giants.
🌟 Professional Presence & Mentorship: While the startup's collaborative culture valued all contributions, establishing credibility through cross-functional engagement and relationship-building strengthened design impact.
🌟 Proactive Communication: Working in a small-team, self-directed capacity required intentional communication—sharing progress, requesting updates, and asking questions to uncover collaboration opportunities. Overcoming initial hesitation with direct inquiry strengthened team alignment and revealed colleagues' openness to transparent collaboration.
References
Building Effective AI Agents. (2024, December 19). Anthropic. https://www.anthropic.com/engineering/building-effective-agents
Create and edit a functional prototype or web app. (n.d.). Figma Learn - Help Center. Retrieved December 3, 2025, from https://help.figma.com/hc/en-us/articles/31304485164695-Create-and-edit-a-functional-prototype-or-web-app
Cui, M., Manyika, J., Bughin, J., Dobbs, R., Roxburgh, C., Sarrazin, H., Sands, G., & Westergren, M. (2012, July 1). The social economy: Unlocking value and productivity through social technologies | McKinsey. McKinsey & Company. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-social-economy
Guide to the Figma MCP server. (n.d.). Figma Learn - Help Center. Retrieved December 3, 2025, from https://help.figma.com/hc/en-us/articles/32132100833559-Guide-to-the-Figma-MCP-server
Introducing Gemini Enterprise. (2025, October 9). Google Cloud Blog. https://cloud.google.com/blog/products/ai-machine-learning/introducing-gemini-enterprise
Microsoft 365 Copilot. (n.d.). What is Microsoft 365 Copilot? Microsoft Learn. Retrieved December 3, 2025, from https://learn.microsoft.com/en-us/copilot/microsoft-365/microsoft-365-copilot-overview
STAMFORD, C. (n.d.). Gartner Forecasts Worldwide End-User Spending on GenAI Models to Total $14.2 Billion in 2025 [07/10/2025]. Gartner. Retrieved December 3, 2025, from https://www.gartner.com/en/newsroom/press-releases/2025-07-10-gartner-forecasts-worldwide-end-user-spending-on-generative-ai-models-to-total-us-dollars-14-billion-in-2025
References
Building Effective AI Agents. (2024, December 19). Anthropic. https://www.anthropic.com/engineering/building-effective-agents
Create and edit a functional prototype or web app. (n.d.). Figma Learn - Help Center. Retrieved December 3, 2025, from https://help.figma.com/hc/en-us/articles/31304485164695-Create-and-edit-a-functional-prototype-or-web-app
Cui, M., Manyika, J., Bughin, J., Dobbs, R., Roxburgh, C., Sarrazin, H., Sands, G., & Westergren, M. (2012, July 1). The social economy: Unlocking value and productivity through social technologies | McKinsey. McKinsey & Company. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-social-economy
Guide to the Figma MCP server. (n.d.). Figma Learn - Help Center. Retrieved December 3, 2025, from https://help.figma.com/hc/en-us/articles/32132100833559-Guide-to-the-Figma-MCP-server
Introducing Gemini Enterprise. (2025, October 9). Google Cloud Blog. https://cloud.google.com/blog/products/ai-machine-learning/introducing-gemini-enterprise
Microsoft 365 Copilot. (n.d.). What is Microsoft 365 Copilot? Microsoft Learn. Retrieved December 3, 2025, from https://learn.microsoft.com/en-us/copilot/microsoft-365/microsoft-365-copilot-overview
STAMFORD, C. (n.d.). Gartner Forecasts Worldwide End-User Spending on GenAI Models to Total $14.2 Billion in 2025 [07/10/2025]. Gartner. Retrieved December 3, 2025, from https://www.gartner.com/en/newsroom/press-releases/2025-07-10-gartner-forecasts-worldwide-end-user-spending-on-generative-ai-models-to-total-us-dollars-14-billion-in-2025