7 Things I Wish I Knew Before Using AI in Product Design
You open Figma, stare at a blank canvas, and spend the next two hours tweaking button radii and debating font pairings instead of actually designing anything. Sound familiar? The painful part is that AI was supposed to fix this — and somehow you're still stuck in the same loop.
I've been designing digital products for a while now, and when AI tools started flooding the space, I genuinely thought I'd cracked the code. Faster workflows. Better output. Less second-guessing. What actually happened was a mess of half-baked prompts, generic outputs, and a growing suspicion that I was using these tools completely wrong.
Here's what I've figured out since then.
The Reason Most Designers Are Getting Mediocre AI Output
The default assumption is that AI is the problem when the designs come out looking generic. It's not. The problem is almost always the input.
When you treat AI like a vending machine — drop in a vague prompt, expect a polished result — you're going to be disappointed every single time. "Design a clean SaaS dashboard" is not a brief. It's a suggestion. And AI tools are incredibly literal. They'll give you exactly what you asked for, which is almost never what you actually needed.
The shift that changed everything for me was treating AI like a junior designer who needs real context. Tell it who the user is, what problem they're solving, what they're feeling when they land on that screen, and what action you want them to take. Specificity is the skill here, not the tool.
Once you nail this, the outputs get dramatically more useful — and that naturally leads to the next challenge.
Prompting for Design Is a Skill Nobody Teaches You
Nobody told me prompting was going to be its own discipline. I assumed you could just describe what you wanted and get something usable back. That's technically true — it's just that "usable" is a very low bar.
What actually works is layering your prompts. Start with the user context. Then the design goal. Then the constraint. Then the tone or aesthetic. Think of it like a creative brief compressed into two or three sentences.
Here's a real example. Instead of prompting "create a landing page layout for a fitness app," I now write something like: "Design a mobile landing page for a 28-35-year-old woman who's tried multiple fitness apps and quit. She's skeptical. The page needs to build trust immediately and get her to start a free trial — not read marketing copy. Minimal, warm, no stock photo energy." The difference in output quality is night and day.
If you want to shortcut this learning curve, this AI prompt pack for product and content creators is something I've used to build out templated prompts for different design scenarios — saves a lot of trial and error.
The prompt skill stacks with every tool you use. Which brings up an important question about which tools are actually worth your time.
Not Every AI Design Tool Deserves Your Attention
The market is noisy right now. Every week there's a new AI design tool promising to 10x your output. Most of them are demos with a sign-up form attached.
The tools I've found actually worth building into a workflow: Midjourney and Adobe Firefly for visual ideation and moodboarding, ChatGPT or Claude for UX copy, user flow logic, and rapid research synthesis, and Figma's AI features for component suggestions and auto-layout logic. Uizard is solid if you're working with non-designer clients who need to see concepts fast.
What none of these tools do well — and this matters — is replace design thinking. They're output accelerators, not problem solvers. If you don't understand the user problem clearly, AI will just help you build the wrong solution faster.
Don't collect tools. Pick two or three and go deep on them until you hit their actual limits. You'll learn more that way.
The Workflow That Actually Saves You Time (Not Just Looks Like It Does)
Here's the honest version of AI in product design workflows: it speeds up certain phases dramatically and adds almost zero value to others.
The phases where AI genuinely helps — ideation, moodboarding, copywriting for UI, generating user personas, and synthesising user research data. These used to eat half a day. Now they take an hour if you're working with decent prompts.
The phases where you're better off just doing the work yourself — final component decisions, accessibility review, anything that involves real user feedback, and the actual thinking behind information architecture. AI will give you an answer here, but it'll be confident and wrong more often than you'd like.
My current workflow looks like this: I use AI in the first 20% (research, ideation, early concepts) and the last 20% (copy refinement, variant generation, documentation). The 60% in the middle is still human work. That balance has kept quality up without killing the speed gains.
The Monetisation Angle Nobody's Talking About
This is the part that most design articles skip entirely, and I think it's a mistake.
If you're a freelancer or building any kind of side income around design, AI fundamentally changes your economics. You can take on more clients. You can productise your process. You can create design templates, UI kits, or component libraries faster than ever — and sell them.
I know designers who have packaged their AI-assisted workflow into Notion templates and prompt libraries and are selling them on Gumroad for $15-$40 a pop. Not life-changing money on its own, but compounding over time with the right audience? That's a real revenue stream built around skills you're already developing.
If that kind of thing interests you, check out the creator and freelancer resource pack at IncomeEdgeHQ — there's some good frameworks in there for turning skills into sellable products without overcomplicating it.
The design skills and the business skills start to overlap more than you'd think once you're inside AI-assisted workflows.
The Quality Control Problem Nobody Warned Me About
Here's something nobody said out loud when AI design tools started trending: the output can be dangerously convincing.
I mean that sincerely. AI-generated UI looks polished at a glance. It has clean spacing, reasonable colour contrast, and sensible component placement. The problem is that it often fails the one test that matters — actual usability by a real human with real context.
I've shipped concepts that looked great on screen and confused beta users almost immediately. Not because the design was ugly. Because the logic was borrowed from patterns that didn't fit the actual user journey. AI replicates what exists — it doesn't always understand why it exists.
Build in at least one round of genuine user feedback before you call anything done. Even five minutes of watching someone interact with a prototype tells you more than two hours of AI iteration. Don't let polished output fool you into skipping that step.
Your Next Step
If you've been using AI in design and feeling like you're not quite getting the leverage you expected, here's what I'd actually do:
1. Audit one recent design prompt you used. Rewrite it with user context, a specific goal, and a constraint. Run it again and compare the output. This single exercise will show you more than any tutorial.
2. Pick one phase of your current workflow and go deeper on AI there. Don't try to overhaul everything at once. If you're slow on UX copy, use AI to draft it for a week and refine your prompting. Small wins compound fast.
3. Consider how your current AI-assisted process could become a product. A prompt library, a workflow template, a Notion OS for designers — these are all things people will pay for. You're building these skills anyway. Start thinking about how to package them before someone else does.
You don't need to master every AI tool on the market. You need to be 20% better at using two or three of them than the average person in your space. That gap is smaller than it looks — and it's absolutely closeable.
Disclosure: This post may contain affiliate links. I only recommend products and services I genuinely believe in. My opinions are entirely my own.
Free Resources
Looking for tools and templates to help you get started? We've put together a collection of free and premium resources over at IncomeEdgeHQ on Gumroad — including checklists, guides and prompt packs to save you time and money.
Top comments (0)