Like a lot of designers right now, I’ve been watching AI creep into every part of the creative workflow—from wireframes to brand voice to literal asset generation. And like a lot of us, I’ve been asking: Could AI actually replace most of what I do?

So I ran an experiment.

Over the course of a few weeks, I challenged myself to replace as much of my design process as possible with AI. Not just for fun—but to really push my thinking around delegation, speed, creative limits, and what still requires human intuition.

What I learned surprised me. Not everything worked. Some things broke spectacularly. But a few tools changed the way I design—permanently.

Here’s what happened.

Step 1: Research & Ideation

I usually start a project with messy Notion docs, Miro boards, and long walks. For this experiment, I started with ChatGPT.

Instead of brainstorming on my own, I fed it raw prompts like:

• “What are the emotional pain points of booking healthcare appointments?”

• “How might we rethink journaling as a multimodal experience?”

• “What are common onboarding UX patterns for productivity apps?”

The results were fast, structured, and occasionally brilliant. I layered in PromptCanvas to break my brief into parts (user, tone, value prop, structure) and it gave me sharper edges to shape around. This combo helped me move 2x faster in the early phase.

What worked: Research, reframing, naming, flow planning

What failed: Lateral thinking. It couldn’t give me wild, illogical, beautiful ideas. That’s still my job.

Step 2: Visual Exploration

This is where things got exciting. I cycled through multiple tools for rapid visual direction:

• Recraft for SVGs, logos, visual branding

• Ideogram for smart text-in-image layouts

• Midjourney for moodboards and atmospheric concept art

• Kittl for polished, editable typographic posters

It felt like having a mini agency in my browser. I used these tools to generate 20–30 visual variations in a single session. I didn’t keep all of them—but they opened doors I wouldn’t have imagined alone.

What surprised me most? I spent less time second-guessing. Having visual options upfront helped me commit faster to a direction.

What worked: Brand direction, color story, graphic systems, style exploration

What failed: UI consistency, system logic, spacing rules. AI doesn’t care about your 8pt grid.

Step 3: UX Wireframes & Mockups

I tried Galileo AI, Diagram AI, and a few Figma plugins to auto-generate wireframes. The output looked… okay. But the moment I needed to customize layouts, introduce microcopy, or embed real logic—it fell apart.

Instead, I shifted back to Figma + ChatGPT. I’d describe screens in plain English and let GPT suggest a component layout. Then I built it in Figma. That flow felt like me designing with a strategist in the room.

What worked: UX copy suggestions, layout planning

What failed: Fully AI-generated mockups—too generic, too rigid, not aligned with user needs

Step 4: Building & Prototyping

I moved to Framer to prototype. It’s not marketed as an AI tool, but I count it here because of how automated the experience feels—live previews, no-code animations, one-click publishing.

I generated some copy with GPT-4, dropped in assets from Recraft and Midjourney, and created a fully functioning prototype in a few hours. The result? A real product feel, not just a figma click-through.

What worked: Fast feedback loops, visual-polish-to-interaction flow

What failed: It’s still not great for database logic or advanced interactions—so it’s not fully replaceable yet.

What I Learned

• AI is best at expanding possibilities, not narrowing them. It helps me ideate and explore—but not make the final call.

• Designers who know how to prompt well will outpace others. Prompt engineering is a skill, not a gimmick.

• You still need human taste. AI can do the work. But only you can decide what’s good, what’s on-brand, what feels right.

What Stayed in My Workflow

• ChatGPT: For early-stage strategy, user insight reframing, and fast drafts

• PromptCanvas: For sharpening project briefs and thinking clearly

• Recraft & Kittl: For visual variations, branding, and layout play

• Figma + Framer: For interface execution and rapid prototyping

• Midjourney & Ideogram: For mood, storytelling, and creative exploration

What Got Cut

• Full mockup generators (Galileo, Magician)

• Auto-layout tools with no customization

• Anything that tried to replace design judgment with “smart defaults”

Final Thought

AI didn’t replace me—but it did make me faster, bolder, and less precious with my first drafts.

The real power of AI isn’t that it designs for you. It’s that it gets you to a good version faster, so you can spend more time refining what matters.

So no, I didn’t fully automate my workflow. But I did redesign how I design.

And that, I’d say, is the real win.