AI Design tools in early 2025

Like most product designers I’ve spent the better part of the last two years exploring every AI tool available and seeing which ones can help improve what I create as a designer.

I don’t mind incorporating new tools and workflows into my process. I’ve done it many times. But so far AI (and let’s be honest, we’re all throwing around this label that is essentially meaningless) has not risen to the task of helping me push myself as a designer.

I’m not talking about using AI for writing, graphic design, logos, banners, or generating visual assets in general. It can, obviously do that, though with vastly mixed results. I’m more interested in seeing if I can use an LLM to speed up parts of my design process.

Following is a brief review of what I’ve found from another exploration in January 2025.

1. ChatGPT

Recently I designed graphs in a dashboard, pulling in data and laying widgets out based on features our team needs. We’d gone back and forth, and I was in the messy middle of sketching out the ideas. I shared my sketches to ChatGPT and asked if it could turn those sketches into a design.

The result was horrible.

I tried, instead, to ask for feedback on what I’d done and what could be improved. It spat out a bunch of ideas. Some I’d already implemented, some were useless, but one idea I hadn’t considered. I asked ChatGPT to wireframe out the new idea and it absolutely failed.

I used the good suggestion to iterate toward a new sketch and was unblocked to take the design forward. That took a few minutes of my time and was valuable enough that I will repeat it.

Using ChatGPT as a rubber duck, by first verifying that it knows what you’re trying to create, and then using it to improve on my own, seems worth doing.

In addition to design, last fall I tested a bunch of uses with ChatGPT to take my designs and flesh them out in code. The results were helpful. As a former web designer and frontend dev I love quickly seeing how something feel in code. There’s a lot of value here and it’s getting better every month. I haven’t found a use case (yet) that allows me to take the code to production level work without tweaking, but it’s value is tremendous for playing with ideas.

2. UX Pilot

I’ve been working on an onboarding flow for a user submission application. There’s a lot of moving pieces, but I wanted to see how UX Pilot could improve on what I’d done with a single screen. Taking an in-progress concept (some details hidden), I wrote out a prompt to UX Pilot asking to recreate what I’d made, without seeing it.

Please create an onboarding flow for me. Let's start with the first page and see how that looks. It should be split into two columns. On the left column in the header we have a back button, a progress bar, and a close button. Below that we have a question (what project is this?), with two buttons for answers. Below that is a continue button. And in the very bottom is the company logo and contact info for help. In the right column we have a photo background with a testimonial from a customer laid over the text.

The first iteration was shocking. It nailed it. Granted, this is a very standard thing to design, replicated a dozen times over on sites like Mobbin. But still I was happy with the results. I decided to push it further and paid for the upgraded service.

I took a dashboard I’d been working on, and uploaded a sketch, while adding the following prompt.

I'd like a project dashboard. On the left you can see a dropdown to switch projects, and below it you can add an episode, and see navigation for items in the project. This should be on a dark background. On the right is the title of the project, the number of episodes, and a checklist for getting started. Below that is an over graph showing retention drop-off for the latest episode, and comments and feedback for the latest episode.

Three examples came back. The third one was not perfect, but still useful. UX Pilot offers an AI helper to enhance your prompt, I tried that and it cleaned up a few things.

Next, I had a screen I wanted to work on, and my sketch hadn’t taken things far visually, so I wrote out the prompt of what I wanted.

I need an all projects screen. It should have a grid of projects on the right, which each project card showing status (in voting, voting complete, application pending), name, number of comments, and a mini retention graph. To the left is a top level navigation, with "+ new show" as the first item, and the active projects below it, also should have a link to settings and help.

The result was something I could work with, but imperfect. I’d need to riff on this several more times.

I then tested a project that’s up next on my list, a revamp of a screen within a mobile app. Though I’ve been thinking through ideas for this, none were at a stage where wireframes or sketches were worth sharing. So I started by uploading a screen of the current experience and asking for ideas to improve it.

Though the previous stages have been helpful, and I can already tell they’re valuable enough to test in my design process, I was curious if UX Pilot (or any LLM) could help me in the thinking stage, when I’m wrestling through a problem.

I uploaded the first view of an existing screen and started with a simple prompt.

I'd like to explore some alternatives to this screen. How can we better present the information shown on this tab?

My entire browser froze at that point and crashed, so I tried again. The result was not what I wanted. It seemed to have just copied the existing design, poorly. When I asked UX Pilot to enhance my prompt, things got weird.

I already have ideas for I want to change, but was curious if UX Pilot could give me new ideas.

I tried a different direction, removing the screenshot and describing the problem I wanted to solve. Again, this was a first attempt and not where I want to go, but a start. The app switched me to desktop mode, so I had to switch back to mobile.

Some of the results that came back were interesting, but not helpful.

I tested editing a screen. Going back to an earlier generation, I selected the featured image widget and asked UX Pilot to turn it into a proper carousel. After a discomforting error, the result was more accurate, which I appreciated for the fact that it could correct something it got wrong from my prompt, but not quite what I had in mind.

Generating screens typically took 10-30 seconds. Not horrible, far faster than I could generate on my own, but improved speed could help reduce frustration with obvious hallucinations and errors.

I also tested the UX Design Review feature with an existing screen, and got back some helpful ideas, several of which are valid and have been brought up as questions in user interviews. The predictive heatmat feature was also interesting, showing where people are likely to look, but I have no way of validating it easily.

This app has enough value to continue pushing and testing in the coming months. It’s not valuable enough to speed up my design process, but it’s a side quest worth exploring.

3. Uizard

Uizard promises a lot. I tested a prompt.

Generate a dashboard. On the left is  account level navigation. On the right is an account status notification and below that widget cards showing each project. Each card should have project title, project status, usage data, and number of comments.

The best result was useless. This is one app I chose not to upgrade to paid as the value seems to be based on throwing Dribbblesque screens in your face.

5. Gemini

Not useful. I tried a bunch of ways to get it to help me with a design, but may have just been doing something wrong.

6. Claude

I asked Claude to review my previous sketch and see what it could do with that. The result was good. Not readily useful, but helpful.

----

Other products seem interesting, such as QoQo, Miro AI, and Relume. Locofy in particular is incredibly interesting. Bolt is scary good. But these aren’t solving the problem I’m interested in right now, which is speeding up my process of design thinking. I’m also willing to throw out my process and shift to something else entirely, such as the Bolt and Locofy taking ideas straight to code to feel them in close to realtime. But I think there’s value in holding to my process a little longer.

Going forward I suspect that I’ll use a mix of UX Pilot, Bolt, Claude, and ChatGPT in my tests to either change my process or improve areas of it.