We Tested Three AI Design Tools on a Real Product. Then a Designer Took Over. Here’s What Happened.

Algorithms can draw a stunning interface in seconds, but they can't sit in a distribution center. Here’s why a perfect prompt still can't replace raw human judgment.

Every few years, a new design tool arrives with the promise of changing everything. I have watched this happen. Each time, the excitement arrives first, while a real understanding of what the technology can and can’t do takes much longer. 

This time, the wave was AI.

AI design tools were starting to show up everywhere; I made a conscious decision not to judge them by demos or social media posts. I wanted to understand what they were actually capable of.

So I started using them on one of our real projects, ID Fresh Food, a company that delivers hundreds of thousands of perishable food packs every single day to retailers across India and the Middle East. A product built around complex human behaviour, used daily by salespeople, distribution executives, and finance teams across different roles, contexts, and levels of digital literacy, where a misplaced design decision has real consequences. 

I put Google Stitch, Figma Make, and Claude Design through that test. The results were more instructive than I expected. 

These tools are incredibly fast. They can give you a screen in minutes and give you a solid place to start. But that’s only the starting point. The rest of the process, like understanding users, making trade-offs, shaping interactions, and refining every detail, is still the designer’s craft. 

A tool that looks great at first glance doesn’t necessarily approach the problem the way a designer would. 

The Three AI Design Tools I Put to the Test

I tested three tools that are getting a lot of attention in the design community today and gave each of them the same brief. I gave all three the same design brief, the same moodboard, the same brand colours, the same visual references, and later, the same wireframe for the ID Fresh Foods sales module. 

That way, I could focus on how each one responded to the same design challenge. 

Google Stitch

Caption: Google Stitch translated the brief into an impressive mobile sales dashboard UI

Google Stitch was the first AI interface design tool I tested. The task was to generate the home screen for a salesperson’s mobile app, a screen that had to communicate priorities at a glance, perform in low-connectivity environments, and feel immediately intuitive to someone checking it for just fifteen seconds before starting their route. 

Within a few minutes, it generated a complete interface. 

Across every iteration, I noticed the same patterns. 

• The visual hierarchy didn’t clearly guide the salesperson’s attention.
• The most important information wasn’t immediately obvious.
• The spacing and alignment needed further refinement.
• The components were placed correctly, but the relationships between them didn’t feel intentional.
• The typography didn’t establish a clear reading flow. 
• Headings, labels, and data existed, but they weren’t helping users scan the dashboard naturally.
• Accessibility considerations such as colour contrast, touch target sizes, and readability still required manual review.

Stitch did a good job of translating my brief into a screen. As I reviewed the output, I realised the decisions that shape a clear and intuitive experience still had to come from the designer. 

Need to validate your AI-powered product? → Explore Aufait UX Usability Testing Services and uncover usability issues before they reach your users. 

Figma Make

Caption: Figma Make generated sales dashboard concepts within the same canvas 

For Figma Make, I used the same wireframe, moodboard, brand colours, and reference material that I had prepared for the ID Fresh Foods sales module. Since everything happened inside the same canvas, it was easy to generate, review, and iterate on different directions. 

While reviewing the screens, I noticed many of the same things I’d seen with Stitch. 

• The interface looked visually complete.
• Information hierarchy still needed refinement.
• Layout decisions didn’t always reflect how people would use the product.
• Brand consistency and accessibility required careful manual review.

As I continued refining the interface, I found myself making the same product decisions I make in every project, like prioritising information, improving hierarchy, simplifying the layout, and shaping the experience around the user. Those decisions still came from the designer. 

Claude Design

Caption: Claude Design generating a mobile sales dashboard UI

Claude Design approached the AI product design process differently. Instead of relying on a single prompt, it encouraged an ongoing conversation where I could describe the screen, review the output, and continue refining it through follow-up instructions.

Before testing it, I made sure the comparison stayed fair. Claude Design received the same design brief, the same references, and the same expectations as the other tools.

I evaluated every iteration using the same approach as the other two tools. 

• I looked at how information was prioritised.
• I checked whether the layout supported the user’s task.
• I reviewed the typography, spacing, accessibility, and overall consistency.
• I asked whether the design reflected the product, the brand, and the people who would eventually use it.

Working through the design conversationally made refinement easier, and the responses improved with better prompts. 

How I Designed the ID Fresh Sales Dashboard from the Ground Up 

Caption: iD Fresh Foods sales dashboard on a mobile device 

The most valuable part of designing iD Fresh Foods was comparing AI-generated screens with the way my team had approached a real product.  

We didn’t begin with Figma. We didn’t begin with a prompt either. We began by understanding what a salesperson needs throughout the day. 

My design team had done deep ethnographic research for this, shadowing sales executives on their daily routes, sitting inside distribution centres, and mapping workflows from stock arrival to payment reconciliation. We knew exactly what each role needed and why. The design challenge was real and layered.

The fastest way to waste time is building the wrong product. Start with user research. → Explore Our UX Research Services

Putting the AI Outputs to the Same Standard 

To be fair, the current generation of AI design tools does certain things remarkably well. 

I intentionally kept the wireframe out of the first test. I wanted to see how well the tools could understand the product from the moodboard, brand colours, visual references, and design brief alone. 

The first output looked structured and visually complete, but it didn’t reflect the priorities our research had uncovered. 

So we tried again with a wireframe. We fed the tool a rough structural sketch, enough to show intent. The output was better. The tool responded to the spatial logic we’d given it. But the gap between “better” and “what this screen actually needs to be” was still significant.

And notably, prompting well takes real time. Writing a prompt that captures operational context, user psychology, connectivity constraints, and information hierarchy is not a quick task. 

Even after multiple rounds of refinement, we still couldn’t reach the quality we achieved directly in Figma, using everything we had learned from our users. 

Going Back to the Design Process 

So we went back to Figma. And the work we did there, grounded in what our research had revealed, was the design that actually made sense for ID Fresh.

We mapped the workflows, identified the decisions users made at different stages of the day, and organised the information around those moments. Our UI UX design process gave us confidence that every decision reflected real user behaviour. 

From there, we built the dashboard around those insights.

• The wireframes came from those conversations.
• The layout came from those workflows.
• The hierarchy came from understanding what deserved attention first and what could wait.

Building the Dashboard Around the User 

When we moved into visual design, we already knew why every section existed. The moodboard helped define the visual direction. The design system gave us consistency. Every review became a conversation about whether the interface reflected what we had learned from the field. 

Looking back, that’s what I’m most proud of. 

Every card, every colour, every piece of information, every bit of spacing had a reason for being there. That’s the part of the process I couldn’t recreate with AI.

The interface was only the outcome. The real design happened long before the first screen was ever drawn.

Where AI Design Tools Still Fall Short 

Here’s where I want to be specific, because vague critiques of AI tools tend to age badly. The limitations I’m describing are structural. They are not temporary quality gaps that better models will automatically resolve.  

• Understanding users 
  AI design tools work from patterns. It doesn’t understand your users’ goals, behaviours, frustrations, or day-to-day context.
• Interpreting research 
  AI can summarise interviews and usability findings. Turning those insights into meaningful design decisions still depends on the designer.
• Understanding context 
  Every product has different users, markets, workflows, and constraints. AI doesn’t naturally understand those nuances.
• Making design decisions 
  Prioritising information, simplifying complexity, balancing user needs with business goals, and knowing what to leave out are still human decisions.

How I Think Designers Should Work with AI

Using these tools on real product work gave me a practical understanding of where they fit. 

I would recommend every designer use it. AI adds speed to the workflow, while designers bring the judgment that shapes the final product. 

Let AI Help You Get Started

Today’s AI tools for designers are incredibly useful when you’re staring at a blank canvas.

Use it to:

• Generate rough layouts and user flows.
• Explore multiple design directions.
• Speed up repetitive design tasks.
• Visualise early concepts before investing time in refinement.

It helps you build momentum, and that’s valuable.

Let Designers Make the Product Decisions

Once the direction is clear, that’s where the designer needs to lead.

That’s the stage where you:

• Prioritise information.
• Shape the user journey.
• Make trade-offs.
• Build visual hierarchy.
• Refine interactions.
• Balance user needs with business goals.
• Design for accessibility, context, and real-world use.

Those decisions define the product far more than the first screen ever will.

AI Can Generate Screens. We Help You Build Products. 

If this article resonated with you, you’re probably asking the same question many product teams are asking today.

How do we use AI without compromising the quality of the product?

That’s exactly how we approach it at Aufait UX.

We use AI design tools where it accelerates the UX design process, exploring ideas faster, visualising concepts, and reducing repetitive work. The decisions that shape the product still come from research, strategy, and a deep understanding of the people we’re designing for.

Whether you’re building a new product, redesigning an existing platform, or exploring how AI fits into your design workflow, we’ll help you separate speed from strategy and build experiences that work in the real world.

If you’re looking for a UX partner that combines AI with research-driven product thinking, let’s start a conversation.

The best products aren’t built by choosing between AI and designers. They’re built by using both where they create the most value.

🔔Follow Aufait UX on LinkedIn for strategic insights grounded in real-world product outcomes. 

Disclaimer: All images belong to the rightful owners! 

Frequently Asked Questions