The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
Most AI products don't fail because of the AI.
They fail because the user experience creates friction.
Here are 5 UX mistakes I keep seeing in AI SaaS products:
❌ Overwhelming dashboards with too much data.
❌ AI outputs without explaining why they were generated.
❌ No loading states or feedback during AI processing.
❌ Generic chat interfaces that ignore the user's workflow.
❌ Poor onboarding that leaves users guessing.
Instead, I focus on designing AI products that:
✓ Build trust with transparent AI decisions.
✓ Reduce cognitive load.
✓ Help users complete tasks faster.
✓ Scale with enterprise workflows.
✓ Turn complex AI into a simple product experience.
Great AI deserves great UX.
What's the biggest UX mistake you've seen in AI products recently?
#AI #SaaS #ProductDesign #UXDesign #UIDesign #DashboardDesign #Figma #Startup #EnterpriseSoftware
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Trending
Claude
Claude has entered the design space. How are you using Claude Design?
Contra University
Learn from expert creatives how to earn more using next-gen AI tools.
creativeaiflow
Creative AI workflows are evolving. What tools do you use, and what are their strengths and weaknesses?
freelancerlife
Freelancer life is wins, pivots, and everything in between. What’s yours right now?
Related posts
Hidden Space is a workspace for creative people working with AI — projects, image and video generation, reusable elements and presets, and a dashboard that shows where your time actually went.
Two days before the challenge showed up I'd been messing with the rawest possible version of this idea — vibe-coding with ChatGPT late at night, no system behind it, mostly just checking whether the thing in my head could exist at all. I wasn't planning anything.
Then the Flowstep challenge landed on Contra, and the timing was almost annoying, because at that point I couldn't pretend I wasn't going to try.
So I thought, fine. Let's do this one properly.
Somewhere around the second prototype it stopped being a submission. I'd wanted this tool for about two years and never had the skill-stack to build it, and suddenly I was designing it instead of describing it to someone who'd never build it either.
Seven screens later it's live, running on production data. I used it for client work this week.
01 · What it is
Hidden Space is a workspace for creative people working with AI. Projects, image and video generation, reusable elements and presets, a gallery that remembers your settings, and a dashboard that shows where your time actually went.
The whole thing sits on one idea: the platform should stay out of the way of the work. The video walks through all of it, so I won't repeat that here. A few things from the process instead.
02 · The colour question got solved by looking
I've spent years of my career debating colour in the abstract and getting nowhere.
This time I rendered three whole directions — achromatic, violet, turquoise — as full workspace screens with the same composition, put them next to each other, and just looked. Achromatic came out stylish and completely impersonal, which I honestly didn't expect. Violet came alive.
Took minutes. I keep thinking about how much time I've wasted not working that way.
03 · Locking the system before the screens
That's what kept the build from drifting. All seven screens were designed in Flowstep against one design system, same tokens and same accent rule, screen after screen.
One accent colour in three tiers. Solid on the primary action. A faint wash on whatever's active. Flat neutral grey everywhere else.
That rule survived production without bending once, which surprised me more than it should have.
04 · I got something wrong, so I'll say it
All seven screens matched their references and I was ready to call the whole thing shipped. Then a small icon bug pushed me into an actual audit, and I found out that "this screen is rebuilt" and "this screen is migrated" were two different claims, and I'd been treating them as one.
Cost me time. Saved the foundation.
If you're moving fast with AI in the loop, verify the thing you're claiming instead of inferring it from your own changelog.
05 · Four tools, four jobs
Flowstep — the design surface. Editable screens I could keep reworking instead of redrawing.
Claude — product reasoning, and the decisions behind the design system.
Claude Code — the production rebuild.
Codex — a second pass over the implementation as it landed.
Underneath all of it the foundation got rebuilt too, from 1,464 lines of hand-written CSS to TypeScript, Tailwind and owned Base UI primitives running on Cloudflare.
06 · What I'd take from it
I didn't have to become an engineer for any of this. I had to know exactly what I wanted and stay disciplined about the system.
That's the part I'd want other freelancers here to hear. The tool you keep wishing existed is a lot closer than it was a year ago.
07 · What's next
7/7 shipped, and that's the foundation rather than the finish line.
Script agents that carry a script through to a storyboard. Audio in the same workspace. Shared styles and a community library that grows the more people use it.
Thank you
To Contra and the Flowstep team, genuinely. These challenges keep pushing the edge of what's possible, and this one pushed me well past mine.
Flowstep's groundwork is what made the difference — the design guide, screens I could keep reworking, MCP into Claude Code and Codex, iteration measured in minutes instead of rebuilds.
I came for a challenge and left with a foundation. Thank you for that.
Try it yourself
Don't take my word for any of it, go walk around inside it. The login below opens the working product on production data.
And if something in there annoys you, tell me. That's how it gets better.
😎 Dropping more in the comments — screens, the bits that got cut, and a few things that went sideways on the way.
⚠️ Links
Presentation
Hidden Space — the working product. Sign in and explore: projects, gallery, presets, elements, dashboard. Generation and sharing are switched off on this demo account, everything else is open.
Username: Flowstep
Password: Contra
Flowstep screens
Figma — behind the process
The biggest mistake businesses make with AI come and learn.
They think they need more AI tools.
They don't.
What they really need is one AI solution that solves one expensive problem.
I've seen businesses pay for:
1. ChatGPT Plus
2. Claude
3. Gemini
4. Cursor
5. Multiple automation platforms
Yet their teams still spend hours every week on repetitive work.
Why?
Because buying AI tools isn't the same as building an AI workflow.
The businesses getting the best results aren't using the most tools.
They're asking better questions.
Instead of:
❌ "Which AI tool should we buy?"
They ask:
✅ "Which task costs us the most time every week?"
That's where AI creates real ROI.
Whether it's lead qualification, customer support, document processing, internal approvals, or reporting, solving one bottleneck often delivers more value than adding five new AI subscriptions.
That's how I approach every AI project:
Start with the business problem.
Then build the AI solution.
I'm curious...
What's one repetitive task in your business that you'd automate tomorrow if you could?
I'd love to hear how others are thinking about AI adoption.
#AI #Automation #BusinessAutomation #AIEngineering
I like your thoughts. In which domain are you building a product?
They thought skipping UX would save money.
It ended up costing them far more.
A client came to me after launching their product.
Everything looked "good enough."
But users were dropping off. Confusion was high. Conversions were low. And support requests kept increasing.
The problem wasn't the branding.
It wasn't the development.
It was the UX.
The redesign they initially postponed became the very thing that transformed the product.
Here's what changed:
→ Simplified the user flow
→ Reduced unnecessary friction
→ Improved visual hierarchy
→ Created a more intuitive experience
→ Designed with real user behavior in mind
The result?
A cleaner product, a smoother journey, and a design built to convert—not just impress.
The biggest lesson?
UX isn't an expense you delay.
It's an investment that prevents expensive mistakes later.
I've shared the complete redesign process, challenges, and final solution in this Behance case study.
💬 Have you ever seen a product suffer because UX was treated as an afterthought? Share your experience in the comments.
♻️ Repost this if you know someone who still believes "we'll fix the UX later."
