Tempo Labs | Agentic AI Design
MY ROLE
Product Design Lead
Product Strategy
TEAM
Product Manager
Founder
Engineers
DURATION
6 Months - Contract
TOOLS
Figma
Tempo Lab editor
Notion
Adobe Premiere Pro
IMPACT
Our AI Copilot gained significant traction, ranking #3 on Product Hunt! The PRD generation feature successfully shipped, and the handoff to engineering was one of the smoothest we’ve ever had.
FINAL DESIGN & SOLUTION
Teams can simply describe the feature they need, and the AI Copilot helps them brainstorm, refine, and ship ideas quickly. With AI-powered suggestions and seamless collaboration tools, it keeps teams focused on building instead of getting stuck in revisions.
START BY GATHERING PRODUCT REQUIREMENTS
The PRD generation is transparent, explaining why each feature is included. Users can accept the suggestions, manually edit or regenerate the content if needed.
PRD GENERATION
Users manage their project through a structured dashboard, where the primary PRD and feature PRDs are stored. They can edit, add, or remove PRDs as needed.
PRD GENERATION
Users can generate a new primary PRD anytime by clicking the ‘+’ button and prompting the AI. Manual PRD addition is also available, informed by usability testing insights.
TICKET GENERATION
The AI simplifies complex feature PRDs by breaking them into specific tasks. Once approved, these tasks are added to the team’s sprint board for execution.
TICKET EXECUTION
Users can skip the sprint board and start tasks directly in the editor. This flexibility helps startups on tight deadlines, enabling individual contributors to work independently.
SPRINT BOARD FOR COLLABORATION & EXECUTION
Tasks appear as tickets on the sprint board for collaboration, like Jira or Trello. The AI can execute tickets in the Tempo Editor, and once completed, they automatically update as 'done' on the sprint board.
USER RESEARCH PROCESS & PLAN
We explored different types of research to understand the problem better. This helped us learn how AI Agents work, how people currently solve this problem, and what the product management experience is like for founders.
01
Competitor Analysis
We analyzed 10 competitors including Devin AI, Butternut AI, Uizard AI, Quickblox, Midjourney and more. This helped us assess where Tempo Labs fits within the landscape of AI agent capabilities.
02
Competitor Audit
We audited 5 direct competitors - Devin AI, Github Copilot, Butternut AI, Microsoft Copilot and Galileo - to see how copilots enable faster iterations and development.
03
10 Semi-Structured Interviews
We conducted 10 semi-structured interviews, including 5 technical founders, and 5 non-technical founders. Our goal was to understand how founders can use AI Agents in their current workflow.
04
5 Participatory Designs
We conducted 5 participatory design sessions with 3 Non-technical founders and 2 technical founders. This was to get a deeper understanding of how the AI copilot can help with better collaboration and iterations.
USER RESEARCH RESULTS & INSIGHTS
We learned that users value a familiar interface and like to brainstorm big ideas but prefer to launch in smaller steps.
80%
Of interviewees want a clear understanding of what AI is going to achieve and how
Participants included both technical and non-technical founders, and both groups wanted to understand the reasoning behind AI decisions to build trust in the product
70%
Of participants found it easier to collaborate with AI when working on a smaller scale - a feature or a component.
This is because smaller tasks feel more accurate, are easier to track, and allow for smoother iterations, increasing trust in AI.
97%
Existing AI agent products lack easy iteration, tracking, and incremental improvements
Participants noted that AI sometimes produces results misaligned with their vision, and fixing them through prompting isn’t always straightforward.
GOALS
We then established our initial benchmark for what we wanted in our product.
ADDRESSING THE HMW
How might we design a seamless and intuitive workflow for generating PRDs, creating tasks and tickets, and executing features effortlessly with AI?
LOW-FI SKETCHES AND IDEAS
We ideated over potential solutions and user flows. I used sketches to solidify thoughts and draw out how I wanted interactions to work.
USABILITY TESTING FINDINGS & ITERATIONS
I ran a small usability test with 3 Product Managers who became founders. My goal was to evaluate how clear, complete, and easy to understand the different features were.
BEFORE
The dashboard felt too cluttered, unclear, and not very easy to understand or use.
.
AFTER
I redesigned the dashboard with better visual hierarchy, clearer AI prompt search text, and more intuitive card design making the "Try Example" feature feel more natural with less explanation.
BEFORE
The PRD takes up the entire screen, distracting users from their main task.
AFTER
To fix this, we designed the PRD as a collapsible side nav to keep users focused on their task.
BEFORE
Previously, PRD generation and the sprint board were part of the same dashboard, but since they are separate tasks, this felt overwhelming and unfamiliar to users.
AFTER
We solved this by creating two separate dashboards, each focused on its task while clearly showing their connection.
BEFORE
It looks like the entire PRD in a task should be executed by AI, but users prefer AI to handle smaller tasks instead.
AFTER
We addressed this by creating a subtask list where users can select specific components for AI execution, improving clarity and information hierarchy.
BEFORE
The tags were unclear and confusing, and the toggle button for task execution felt too significant for a simple toggle.
AFTER
We redesigned the card to display essential details like the associated PRD, due date, priority, and subtasks, ensuring task execution happens only when users dive deeper into the task card.
CREATING DESIGN SPECS
We created a detailed design spec for handoff, making it easier for engineers to develop seamlessly and ensuring a smooth handoff process.
THE FINAL SCREENS
After bringing everything together, we designed a 0-1 AI Copilot to help founders collaborate, brainstorm, and refine their product more effectively.
REFLECTION
This project was an incredible learning experience, especially working in a fast-paced startup environment.
Embracing Pivots – Navigating multiple pivots taught me how to adapt quickly without compromising time or quality.
Communication is Everything – Aligning visions with the founder and collaborating with engineers was crucial, especially for such a technically demanding product. My computer science background definitely came in handy!
Smooth Handoffs – I learned how to create beautiful and intuitive design specs for engineers, leading to one of the smoothest handoffs ever.
Deepened Passion for AI – This project fueled my excitement for designing AI-driven products, and I can’t wait to build more!
