There’s a particular part of the design process that can take more time than it should. It’s not necessarily solving the problem. It’s getting started.
You know the brand.
You know the existing website.
You know the design system.
You know what the client is asking for.
Maybe you need to add a new page, rethink an existing layout, or create a few variations of something that already exists. But there’s still that blank Figma frame staring back at you.
This is where Figma’s AI agent has become an interesting addition to my workflow. Not because it can replace me as a designer. It can’t. But because it can sometimes get me from “Where do I start?” to “Okay, now I have something to work with” much faster. And I’ve found that distinction matters.
AI is more useful when you aren’t starting from scratch
When we think about AI design tools, it’s easy to imagine giving them a prompt and getting an entire finished website in return. That’s not really how I see the Figma AI agent being most useful. For me, its biggest value comes when something is already established.
An existing website. An established brand. A design system. A set of components. An existing page that needs to be reworked. A new page that needs to fit into an existing experience. The more context there is, the more useful the starting point becomes.
If I’m designing something completely from the ground up, there are a lot of decisions that need to happen before AI can meaningfully help: What is the visual direction? What is the brand personality? How should the information be organized? What should the experience feel like? Those are design problems, not just production problems. But if I’ve already made many of those decisions, AI becomes much more interesting. Instead of asking it to invent an entire design language, I can ask it to work within one that already exists. That’s where I think the real time savings start to show.
Think of it as a first draft generator
One of the most useful ways I’ve found to think about the AI agent is this: it doesn’t need to give me the answer. It just needs to give me something to react to. That might not sound revolutionary, but there’s a huge difference between staring at an empty canvas and looking at a design that is 60–70% of the way there.
Once something is on the screen, I can immediately start making decisions. This spacing isn’t right. That heading needs more emphasis. This component doesn’t belong here. The content needs to be reorganized. This section feels too repetitive. The visual hierarchy isn’t working. The colors need to follow the existing system. That image doesn’t feel like the brand.
Those are decisions I can make quickly because I have something in front of me. In that sense, AI isn’t necessarily doing the design for me. It’s helping me start the design conversation with myself.
Where Figma’s AI agent shines
1. Getting past the blank canvas
This is probably my favorite use case.
Instead of manually setting up every frame, section, and placeholder, I can use AI to establish an initial direction and then start designing from there.
The first version doesn’t have to be perfect. It just has to exist.
2. Adding pages to an established website
This is where I think the tool becomes particularly valuable.
Imagine you’re working on a website that already has:
- Established typography
- Defined colors
- Existing components
- Buttons and form styles
- Navigation patterns
- Card layouts
- Spacing conventions
- A recognizable visual language
You’re not looking to reinvent the website. You need to add something that belongs there. That’s a much more constrained problem, and AI can be helpful within those constraints. Rather than starting with a blank page, I can use the existing design as a foundation and have the AI agent help create an initial layout that follows the established direction.
From there, I take over.
3. Exploring variations
Another advantage is speed.
Sometimes I have an idea, but I’m not completely sure which layout will work best. Creating several variations manually can take enough time that you end up committing to your first idea simply because it’s already built. AI makes experimentation cheaper. I can explore a few directions, compare them, and then take the strongest idea forward. The important part is that I’m still the one deciding which direction makes sense.
4. Repetitive design changes
There are also tasks that simply aren’t particularly creative. Updating similar elements. Making bulk changes. Adjusting layouts. Filling out repeated content. Creating variations.
These are the kinds of tasks that can eat up time without necessarily requiring much design thinking. Having an AI assistant handle some of that initial work can free up more time for the parts of the project where design judgment actually matters.
But here’s where I still come in
This is probably the most important part of the conversation. The AI agent doesn’t finish the design for me. And I don’t think it should.
Even when the generated result looks surprisingly polished, I still need to go through it. I need to check the typography. I need to check spacing. I need to check alignment. I need to check components. I need to check colors. I need to check hierarchy. I need to check whether the page actually makes sense. And I need to make sure the new design actually feels like it belongs with the rest of the project.
AI can recognize patterns. A designer needs to understand why those patterns are there. That’s a pretty important distinction.
Sometimes “almost right” isn’t right
One of the interesting things about working with AI is that the output isn’t necessarily bad. Sometimes it’s actually pretty good. It’s just not quite right.
Maybe the spacing is technically consistent but doesn’t create the right visual rhythm. Maybe the hierarchy is logical but doesn’t emphasize the information that matters most. Maybe the layout follows the existing design system but feels too repetitive. Maybe the page looks polished but doesn’t quite feel like the brand. These are the moments where experience matters. A designer isn’t simply checking whether something looks acceptable. We’re asking whether it communicates the right thing, to the right person, in the right way. That’s much harder to automate.
The downside: AI can create more work, too
Of course, there are times when using AI isn’t necessarily faster.
If I have to spend 30 minutes correcting an AI-generated design that would have taken me 20 minutes to create myself, I haven’t saved any time. There can also be inconsistencies that aren’t immediately obvious.
- A component might look right but not actually use the correct component.
- A spacing value might be slightly off.
- A font style might not match the system.
- A section might introduce a pattern that doesn’t exist anywhere else on the site.
These details are easy to overlook when you’re focused on the bigger picture. That’s why I don’t think the AI output should be treated as the finished product. It’s a starting point that needs a designer’s review.
The biggest benefit isn’t actually “designing faster”
It’s reducing the amount of time between an idea and something you can evaluate. That’s an important distinction.
The design process isn’t just about producing pixels. A huge part of the process is evaluating ideas.
- Does this work?
- What about this?
- What if we move this?
- Could this be simpler?
- Is there a better way to organize this information?
Once there’s something on the screen, those questions become much easier to answer. That’s where I think AI has the potential to make designers more efficient without making the designer less important.
So, is Figma’s AI agent worth using?
For me, yes, but with realistic expectations. I wouldn’t hand it a project and expect it to produce a finished, client-ready design. I also wouldn’t use it as a substitute for understanding the brand, the user, the content, or the existing design system. But I would absolutely use it to help me get started. Especially when I’m working on an existing project, adding pages to an established website, or need to explore several directions quickly. And especially when there’s a lot of repetitive production work standing between me and the actual design problem.
The biggest shift for me is that I’m not thinking of AI as the designer. I’m thinking of it more like a very fast design assistant.
- It can put something on the canvas.
- It can give me a direction to react to.
- It can take care of some of the repetitive work.
- It can help me move faster.
But I’m still responsible for deciding whether the result is actually good. And honestly, that’s probably the way I want it.
I’m not handing the design over to AI. I’m giving myself a head start.
