I designed a Setup Framework to support the scale, launch and functioning of AI Agents
Year
2025
Platforms
Web
Role
Research, Interaction Design, UI Design, Prototyping
Team
Product Managers, Designers & Design Managers, Engineering Architects
What started as a single AI Agent to improve dealership efficiency, quickly evolved into a need for multiple AI Agents. Each agent would be responsible to automate a specific part of the user journey.
This kind of scale required a systematic, efficient and plug-and-play model of Setup that would allow customer to get any agent of choice up and running within a few minutes.
The entire process was more collaborative & iterative in nature instead of a linear workflow. I worked closely with Product and Engineering leaders to research, explore design solutions and refine ideas iteratively over the course of a few months. While we started building this framework for a 1-2 AI Agents, as we moved closer to the launch timelines, we tested the framework with more AI Agents.
We started with a 1-week workshop with members of design, product, engineering, data and more teams. The idea was for everyone to understand business strategy, discuss constraints and create a plan.
Based on the kickoff workshop, I designed the first version of the Setup. It was far from perfect but kicked off productive conversations as we finalised the details over the next few months.
I took a pause amidst the constant back and forth of ideas to comprehensively re-look and identify the user needs, look at competitors before resuming designs.
After further clarity, I mapped out the end-to-end user flows, edge cases, error handling and other minute details.
Designed detailed concepts and prototypes based on the user-flows.
Started larger discussions with the design, product and engineering leadership. The feedback was largely positive along with a few updated required in the Information Architecture, edge-cases and other minute details.
Designed and prototyped end-to-end user-flows, edge cases, error handling, UI states and more.
Started onboarding other AI Agents that were in progress to test the robustness of the framework, identify gaps and scalability issues.
The final designs look something like this.
