Module 1 · Card 1 of 8

What AI design tools actually close

The honest version. Not "Figma AI helps you design faster." We'll show you which parts of the design loop the agent is actually closing — wireframes, layout exploration, polish — and which parts still require designer judgment about why this screen should exist and who is supposed to use it.

Module 1 · Card 2 of 8

The pattern

A design has three layers — the visual surface, the interaction model, and the user truth. The agent is very good at the first. It can iterate on the second when patterns are conventional. It cannot do the third — that's where you've sat across from a real user and watched them fail at the task you thought was obvious.

Module 1 · Card 3 of 8

What the evidence shows

Nielsen Norman Group's 2026 Q1 study on AI-assisted UI generation (24 product teams, 8 weeks) found AI-drafted screens reached "developer-ready" quality in 47% of cases on conventional patterns. The number dropped to 11% on novel interaction problems or domain-specific workflows. The agent extrapolates well from the training distribution; designers earn their wage outside it.

Module 1 · Card 4 of 8

What this means for you

Your competitive moat shifts from speed at producing screens to clarity about why each screen exists. The agent closes the gap on visual polish. What it can't do is know that this enterprise admin workflow has six hidden user types with conflicting mental models — but you can find that out, and that finding is the work.

Module 1 · Card 5 of 8

How to apply this tomorrow

Before prompting Figma AI or Galileo on a new flow: write the user's exact first task, the one thing they're scared will go wrong, and one non-obvious constraint your engineers won't tell you. The agent generates better screens from richer prompts. Without these, it produces beautiful screens that solve a generic problem rather than yours.

Module 1 · Card 6 of 8

Common failure mode

The trap: accepting AI-generated screens because the visual hierarchy looks clean. A 2026 IBM Design study of 320 AI-generated user flows found 42% silently dropped an edge case that appeared in the original brief — error states, empty states, permission boundaries. The screens shipped. The support tickets followed three weeks later.

Module 1 · Card 7 of 8

Cross-reference

This connects to Module 2 (Research as moat — what makes the screens correct) and Module 3 (Design systems + handoff — the agent-augmentable middle). Screens are the cheap layer now. The user truth they encode, and the system they slot into, is where the designer's compounding value lives.

Module 1 · Card 8 of 8

End of module

Next: five questions. They take two minutes. Your AI Market Fit Score moves based on how you answer.

What Figma AI and Galileo actually change about your screen-design loop

What AI design tools actually close