A lot is being written right now about how AI will change design workflows and what it means to be a designer in an AI era. Some of it is directionally right. Some of it feels premature. Six months ago this was barely a mainstream conversation. Today it is everywhere. Six months from now, it will likely look different again.
That alone is a useful signal. We are clearly in a transition phase, not an end state.
To understand where things might actually go, it is worth revisiting why Figma won in the first place.
Figma did not win because it invented design. It won because it collapsed a fragmented workflow into a single place.
Before Figma, collaboration meant passing files around or locking them while someone else worked. Prototyping meant exporting designs into a separate tool like InVision. Handoffs meant exporting again into Zeplin or static specs. Each step added friction. Each step increased the chance of misalignment.
Figma brought collaboration, prototyping, and handoff into one shared, instantly accessible surface. Multiple people could work in the same file. Prototypes lived next to the designs. Developers could inspect the same source of truth. Nothing about this was conceptually new. It was just the best packaged version of the entire workflow. And because it solved multiple real problems at once, it became the default. Today, nearly every serious product organization runs on it.
That is an important lens for evaluating AI design tools.
Right now, AI tools absolutely help people move faster. But they still lack several of the fundamentals that made Figma indispensable.
Most AI driven workflows are still multi tool chains. Figma to MCP. MCP to another AI app. AI output to Git. Git to a repo. Each handoff introduces friction. When you are just testing ideas, that friction matters.
There is no true live collaboration. Sharing a vibe coded artifact from Cursor or Claude Code still takes effort. Others cannot easily jump in, tweak it, or explore variations together in real time.
You also have to be extremely explicit. These tools shine when you know exactly what you want and your goal is to get to code as fast as possible. If speed to implementation is the primary goal, that can be a win.
But design is not just about generating code quickly. Design is about identifying the right problem before committing to a solution. It is about de risking development, which is still expensive and time consuming even with AI. Figma succeeded because it supported that early ambiguity. It made exploration, iteration, and shared understanding cheap.
Those fundamental design problems are not fully solved by any AI tool yet.
Figma Make points in the right direction. It preserves the core strengths of Figma like instant sharing and collaboration while layering in generation. But today, at least from my own testing, the output is not yet something you would comfortably ship or build on directly.
Which brings us to the more interesting question.
The end state probably is not another standalone app or another clever plugin. The winning workflow has not existed yet.
If history repeats, the next breakthrough will collapse multiple steps again.
Design vectors and components should be generated as fast or faster than building them by hand. Prototypes should converge with real interfaces, not sit beside them. Sharing should be effortless and live by default. Exploration should be collaborative, not serialized. And the distance between idea, prototype, and usable artifact should continue to shrink without sacrificing the core purpose of design.
We are not there yet.
But if you look at why Figma won, the direction becomes clearer. The tool that truly wins in the AI era will not just make individuals faster. It will collapse the workflow in a way that removes entire classes of friction.
That is still the game. And it is still wide open.