I built tools into my design process to move faster on specific bottlenecks and get better results from research. Three tools, three different bottlenecks.
Early in a project, directions shift and requirements are still taking shape. I started using AI to work through the problem before committing to any one direction.
Before opening a design tool, I'll brainstorm different ideas and use my agentic tools to work through a range of approaches, looking for where the logic breaks or where an assumption doesn't hold.
It changed how I start a project. Instead of committing to one direction early, I can explore more ground in less time and arrive at the design phase with a clearer reason for the direction I chose.
The most straightforward path from concept to a product is through code. I have produced a few artifacts with agentic code:
The difference shows up in reviews and research sessions. People stop narrating the screen and start reacting to the experience. I get feedback that's actually about the design.
Before recruiting for a study, I use AI to probe the research plan for assumptions I haven't examined yet.
A full study takes weeks to recruit, run, and synthesize. By the time it comes back, the design has often already moved. Running that pass first means I go into recruiting with a tighter hypothesis.
Status updates, handoff packages, and decision logs used to take up a meaningful chunk of my time. I replaced most of it with scripts.
The scripts pull from design files and project state and run automatically. The team gets what they need without anyone having to stop and produce it.
Two tools doing most of the work.
Claude Code
AI coding assistant I use to build research tools, prototype for web and mobile, and automate the repetitive parts of my process.
Kiro Crew
Amazon's AI-powered IDE with a multi-agent system. I use it for spec writing, code generation, and running the research prep steps I used to do by hand.