My AI Workflow

I do discovery, design, everything end-to-end - and often open a PR.

Yes, AI is the buzzword right now - so legally I had to include a section about it. This is not a "thoughts on AI" essay. Below are the tools I actually use, day to day, and what each one is good at.

The pattern is the same across every stage: AI helps me move faster on the messy middle - synthesizing notes, exploring directions, prototyping in code - while I keep judgment, taste, and final calls on me.

Chelsey's AI workflow at a glance

How I use AI day to day

01 · Research

Synthesize notes & spot patterns

I record interviews with Granola and pipe transcripts into Claude to surface patterns, sharpen what's worth paying attention to, and pull quotes I'd otherwise miss. I'm also building Claude skills and plugins so the workflow gets smarter over time.

Granola Claude Custom plugins
02 · Ideation

Explore directions quickly

I use a mix of Lovable, v0, Claude, and Figma Make to pressure-test ideas and get past the obvious first solution. Most outputs aren't great - but bad ideas are useful too. They build character. And occasionally, taste.

Lovable v0 Claude Figma Make
03 · Design & build

Prototype directly in the repo

For real polish I clone the repo locally and prototype in code. Claude Code in VS Code and Warp lets me test interactions and raise front-end PRs myself - less handoff, less "that's not what I meant."

GitHub Claude Code VS Code Warp
04 · Tickets

Turn meetings into Jira tickets

With Granola notes from a working session, I have Claude draft Jira tickets - acceptance criteria, edge cases, the works. UI/UX may already be explored in a PR, but the logic still needs to be written down clearly so engineering can ship it.

Granola Claude Jira