Medium: AI in product design — where we are now in 2026

David Robinson published this article in February 2026 as a deliberate reality check against the more breathless predictions made two years earlier. Where many articles from 2024 described AI as imminent designer replacement, Robinson’s take is more useful: AI absorbed the tedious parts of the workflow and forced the discipline to get serious about design systems.

The piece is structured as a personal account from a working product designer. Robinson covers four areas where AI made a genuine difference: design systems, prototyping, tooling interoperability, and the shifting nature of the blank canvas problem.

On design systems, the change Robinson describes is structural. Design systems are no longer built primarily for the humans who implement them. They are built for the AI tools that read them. If component tokens are not machine-readable, if component logic is not documented in a way that an LLM can parse, the design system stops being usable for AI-assisted generation. This is a concrete change that has altered how design system teams write documentation and structure component libraries.

On prototyping, Robinson finds the prompt-to-prototype path genuinely useful for early-stage exploration. Figma Make and similar tools can generate a working wireframe from a description in the time it previously took to pick a font. He is clear that this output is not finished work — it is a starting point. The designer’s job shifts from blank-canvas creation to evaluation, selection, and refinement. That shift is real and saves time, but it also demands that the designer maintain enough craft knowledge to recognize what the AI got wrong.

The interoperability section covers something that was a genuine friction point in 2024: tools that could not share context. By early 2026, Robinson describes AI tools that are meaningfully connected. Generating a FigJam diagram from a Claude conversation, or pushing Figma component data into a coding assistant, now works without manual export steps. The workflow is still not frictionless, but the gaps have narrowed enough to matter.

Robinson closes with a principle that runs through the whole piece: the design teams that thrive are those that know when to generate, when to refine, and when to step in with judgment that machines cannot replicate. The article is well-suited to mid-level and senior product designers who want an honest account of where the day-to-day work stands in 2026.