Something interesting I’ve been thinking about lately is how tools like Figma, Sketch, and Paper will fare in the era of AI coding agents. In this article I’m going to dump some of my thoughts on patterns I’ve noticed in the space and some predictions for the future.
I say “Figma file” a lot in this article but that’s just to reference a vector design canvas (could be made in Paper, Sketch, etc.)
Also, please take what I say in this article with a huge grain of salt. Just observations from a young engineer 🙂.
Out with the old
If you’ve been an engineer for the last several years you’re probably familiar with the old workflow, if not still using it: designer designs, engineer implements. This workflow was great when code was being written by hand. Designers could spend meaningful time mulling over UX and the best way to visually present a certain feature to users. This form of iteration was cheap and fast.
In with the new
The velocity of engineers and product teams has increased exponentially with the introduction of AI coding agents. This makes the old “designers design, engineers implement” workflow breakdown in a few fundamental ways:
Products are evolving so quickly from an engineering standpoint that a Figma file will quickly drift from what’s in production.
Product and non-technical teammates are writing code now more than ever due to these new AI tools.
Due to these changes that AI has made in how we work, keeping a Figma file up to date with what’s in the product is becoming more burdensome than helpful.
The Death of Vector UI Design
Might be a harsh take now, but I think the day is coming when vector UI design will not be around as we know it today. Designers will be needed more in the AI era than ever before but their traditional way of working is changing more rapidly than ever.
Dragging to size a frame and typing in padding values into Figma is now comparable to writing simple backend logic or building a boilerplate heavy React component. Just as AI has abstracted this work from programming, it will continue to abstract it from designing. In this way, vector tools like Figma that don’t adapt will die simply due to the fact that designers are now able to access the codebase more closely than ever before.
Survival
How will these tools survive and evolve? I don’t know.
What’s interesting to me now is some of the ideas the teams building these products are experimenting with given how AI has changed the space.
The Paper team, for example, is building a great MCP tool to give agents access to the vector canvas. The MCP allows an agent like Claude Code to generate content in the Paper canvas, just as if you were building UI code. Paper then allows you to export items on the canvas to HTML which can simply be fed back to a coding agent for implementation and further code iteration.
I think this tool is very interesting but it continues to beg the question, why do we need the canvas at all? This question is probably naive and I’m not best fit to answer it but code has become so cheap and easy to iterate upon via natural language that any non-natural language interface will need to be very compelling in order to win in the future. There is an argument that non-technical designers and PMs should be able to make small surgical changes but as models improve and tools get better, these surgical changes will be easier and easier to make.
All of this is to say, I think these tools will change quite a lot in the coming years and I’m excited to see the future of the space and maybe build something to help designers :)