02 · PRACTICE · NOT A CASE STUDY
AI for Product Designers
How my design workflow has evolved — and what AI has become inside it.
- ROLE
- AI Workflow Exploration
- YEAR
- May 2026
- SURFACE
- Workflow · Practice
- FORMAT
- Reflection + Live demo
01 · THE SHIFT
What changed in two years.
Before Maven’s AI for Product Designers track, my days still moved in familiar beats: FigJam / Figma for structure, a prototype for proof, Slack threads for writing tweaks, engineering for the sharp edges. Useful, but the handoffs between ideation, copy, scaffolding, and build still ate hours.
The coursework pushed me to keep those stages but strip the waits in between — pairing Claude / ChatGPT for language and teardowns, Figma Make for directional layouts fast, Cursor and Lovable when clicking mattered sooner than polish, Google AI Studio whenever a multimodal spike was quicker than mocking it, Webflow when the deliverable truly was publishing. Assignments surfaced where I was over-documenting versus where an outline plus an AI collaborator was enough for the next reviewer.
Below is how I catalogue that rhythm now — five pragmatic passes and a live demo so you can run the loop here, not read about it in the abstract.
Cheap iteration beats perfect first drafts — the loop just has to stay legible across tools.
I still move from question to artifact like any senior IC: frame constraints, ask the model narrow jobs, spike multiple answers, reconcile with accessibility and engineering reality, ship into Figma, Webflow, or code. Prompting is not a replacement for critique — it buys parallel variants so refinement happens on real options instead of hypothetical ones.
03 · A REAL PROMPT
Prompts are design specs.
During the Maven sprints — and afterward on dense enterprise surfaces — I started treating prompts like mini specs: stakeholders, stakes, typography of the answer. Below is a representative excerpt I still reuse when pressure-testing labeling for consequential actions inside admin flows (same structure, edited per feature).
Role: Senior Product Designer improving UX writing for a SaaS dashboard.
Goal:
Help admins understand a high-risk action before removing user purchase access.
Constraints:
- Maximum 13 words per sentence
- Clear and direct language only
- No jargon or overly dramatic wording
Task:
Create 4 button label options for the confirmation modal.
For each option include:
- The button label
- A short explanation of why the wording works
Focus on different communication styles:
- Direct
- Reassuring
- Specific
- Calm and clear
Try it on a real
design problem.
Three canned prompts modeled on Maven assignments and analogous IC jams — UX tightening, KPI empty states, Cursor handoffs. Run one inside the same artifact pane I use to think out loud.
Session limit: 5 runs. Real model call. No chat history kept.
SYSTEM PROMPT IS SET ON THE SERVER. NOT EXPOSED.
FigJam · Figma
Briefs, journeys, critiques with PM partners — unchanged anchor, quicker annotations.
Figma Make
Directional comps when I want layout options before consolidating into the canonical file.
Claude
Longer teardowns and spec drafting when I want a patient second brain with context.
ChatGPT
Fast variant passes — different biases than Claude — great for ripping apart copy tone.
Google AI Studio
Cheap multimodal experiments (audio/visual prompts) without standing up tooling.
Cursor
Where UI intent becomes production-ish code snippets and small Next.js tweaks live.
Lovable
Clickable shells after FigJam narrows IA — helps me sanity-check rhythm before engineering.
Webflow
Launch surfaces that ship without tying up eng; still part of weekly maintenance.
05 · REFLECTION
What this practice is for.
The Maven course helped me stop treating AI like a shortcut and start treating it like part of the workflow. The biggest shift wasn’t speed alone — it was learning where AI actually helps, where it creates noise, and where human judgment still matters most.
A lot of the assignments pushed me to work more openly and iteratively: testing rough ideas earlier, writing clearer prompts, validating assumptions faster, and tightening the loop between thinking, prototyping, and shipping. That changed how I approach design reviews and even how I structure my files.
AI now sits naturally inside my process across research synthesis, UX writing, prototyping, case-study framing, and implementation exploration. Not as a replacement for craft — more like an active collaborator that helps me move through ambiguity faster while staying intentional about decisions.