Human judgment first. AI acceleration second.
Case Study
The Challenge
Design is a discipline that improves with friction — pushback, alternatives, a second opinion. Working solo means those feedback loops are either missing or slow. You can spiral inside your own assumptions without anyone to object.
AI doesn't replace a team. But used deliberately, it can recreate the structure of collaborative thinking: proposing alternatives you wouldn't have considered, surfacing objections before a stakeholder does, and compressing research synthesis from days to hours. The challenge is keeping the human judgment at the center — not outsourcing it.
The Process
Three scenarios that show what this actually looks like — not as theory, but as a working method.
Five user interviews. Thirty pages of notes. I feed the raw patterns to Claude and ask for synthesis — not conclusions. The prompt is: "What tensions or contradictions exist across these responses?" The AI surfaces the friction. I decide what it means.
I pitch a flow direction and ask for three alternatives, then ask for the strongest objection to my original idea. The pushback — not the alternatives — is where the real thinking happens. Having to defend a direction, or abandon it, sharpens both options.
I write the design rationale in plain language — why this decision, what was traded off, what would break this. AI restructures it for a handoff doc. Faster, cleaner, still mine. The thinking happened before the prompt.
Four tools, each with a defined role. The discipline is not mixing them — keeping each bounded to what it's actually good at.
Long-context reasoning and research synthesis. Used for compressing interview notes, structuring problem frames, and stress-testing design rationale. Strongest at nuance and sustained argument.
Fast generation of alternatives, copy variations, and quick-turn lookups. Used when the goal is quantity over depth — get five versions of something quickly, then apply judgment to select and refine.
Where the actual design decisions are made and tested. AI informs what goes into Figma; Figma is where the conclusions are drawn. Not a place where AI makes choices — a place where the designer does.
Research capture, prompt logging, and synthesis documentation. Tracking what was asked, what was useful, what failed — building an institutional memory of the AI-augmented process over time.
The Solution
The line between what AI does and what the designer does isn't ambiguous — it's deliberate. Letting that line blur is where AI augmentation becomes AI dependence.