Figmaless design flow. New approaches in the Era of Disruption.
For many years product design has followed the same trusted playbook: first research, prototypes last. Designers researched, explored problems, developed ideas and delivered solutions. That process was then followed by a handoff to developers who translated that into code. A thoughtful, deliberate process.
That was The Playbook and The Order of things before the GPT moment. The phrase “Figmaless design flow” started appearing in the social space, sounding very radical. Figma is by far the most used tool for first stages of product design. So skipping the tool that dominates the bulk of designers’ workflow?
But the shift wasn’t about skipping Figma at all. It was GPT making code generation so easy that the initial design stage could be automated, giving the illusion that it was optional…
Flipping the table
To understand what is actually happening I started looking at AI tools for product development. Turns out there are so many of them. Developing a product in tools like these comes down to a prompt window and ends up with a clickable product minutes after you hit enter. The first time is mind blowing, because for years, it took weeks of work to get to that clickable moment. What once came last now comes first. The settled order described before is disrupted by AI. These tools generate both code and UI components fusing design and development together.
Figmaless design flow?
Figmaless sounds catchy, but the change we are experiencing is a reversed strategy, rather than eliminating Figma from product development. Generative AI is probably the biggest disruption in the long established design flow. In the traditional course it was nonsense to waste time on building something that won't address user problems. But gen AI operates differently. You start with a working product and then trim it down, sculpt like a stone (subtracting) rather than gradually adding new material.
Vibing tools
The ecosystem is buzzing with tools supporting the reversed strategy: Framer Ai, Replit, Lovable, Bolt.new, Google Stitch, Magic Patterns, V0 to name a few… They blur the line between UI and frontend because both are results of vibe coding. In vibe coding you describe the product and AI generates both design and code. I tested the same simple idea of a responsive to-do list component using some of these tools. The results illustrate the strengths and limitations of an AI-first approach.
Democratizing product design
Introduction of generative AI in product development opened doors to people from outside the field. Suddenly, people without years of design training can generate interfaces and working code. So if anyone can do it, what becomes the value of design? At the same time, the new flow comes with clear trade-offs in how products are researched, and in how code is written. Which also raises a general question on the product’s overall quality.
The bottleneck of AI-first design
From a design perspective, the research is often shallow or nonexistent. It is tempting to experiment by jumping straight into the prompt editor and starting from there. But weak research most of the time gives a weak product. The outputs often share a similar aesthetic, unless you describe very precisely what you want, which ironically means going back to the original playbook. As we start with an abundance of generated components we face the challenge to discard the unnecessary features. An embarrassment of riches.
From a code perspective, it’s easy to end up with a messy code base. The interfaces may look polished, but under the hood the code is often inconsistent, lacking structure, naming discipline, or scalability. Few tools yet handle design systems well (and unless you are building a website, chances are you do need a design system for your product). It also raises questions regarding safety. There are already reports of situations where an AI agent wiped out the whole code base, or exposed personal data.
Steve Jobs on Artificial Inteligence, Aspen 1983
Where AI really fits
AI-first tools are made with a mission to remove every possible friction in software development. It expands what's technically difficult, but simply patching everything with AI won’t make products better. Building something worthwhile still depends on empathy, research, and solving real problems for real people. We can successfully use AI-tools to build the logic of understanding things, not to produce the ready solution.
That’s why I don’t see AI as a threat to designers or to Figma. If anything it strengthens the profession’s position in business and strategic operations. It also highlights the parts AI can’t replace: research and decision-making. The AI-first approach is less as a replacement, and more as a new entry point. It works brilliantly for prototyping, validation and testing, when speed matters more than polish. But it doesn’t erase the need for design thinking.
