Most design teams are somewhere in between on AI. Clients want it on everything. Designers are split on using it at all. The team calls its stance "curious" and moves on.
That in-between spot is not a lack of opinion. It is a lack of framework. AI did not close the design skill gap. It moved it, from execution to judgment.
The floor rose. The ceiling did not.
The tools got good enough, and "good enough" became everyone's starting point almost overnight. A competent, forgettable layout now takes seconds. UX Collective writer Matej Latin argues that 90% of designers risk becoming unhirable, not for lack of talent, but because they are still solving problems the old way.
When everyone ships from the same floor, shipping the floor gives you no edge. The ceiling only moves when a person decides where it should be.
Three zones: automate, protect, decide
Every AI design decision falls into one of three zones. Sort your process into them before you sort your prompts.
- Automate. Layout variations, resizing, first-draft copy, asset generation. AI runs without a human checking every step.
- Protect. Point of view, visual identity, what makes your brand yours. Never hand these to a prompt.
- Decide, case by case. Is the prototype close enough to ship? Is the AI's version actually better? A human reviews, every time.
Taste moves away from AI
When anyone can produce the same flat, generic layout in seconds, that layout stops standing out. Designer Michal Malewicz describes big brands moving toward glass, metal and hand-drawn interfaces, a direction he calls Skeuomorphism 2.0. It exists because AI makes testing a hundred directions an afternoon's work.
Before you ship an AI-generated layout, ask: could your direct competitor ship the exact same screen? If yes, the work is not done.
The job moves into the real product
Designers who work in the live product, not just the deck, make faster and better decisions. A designer at Amplitude who had never shipped code now owns much of the frontend, because polished prototypes could not keep pace with daily releases.
Signs your team is still stuck in the deck: your best decisions get made in slide review, nobody can open real customer data without asking, and "ship it and see" takes longer to schedule than to build.
The skill that matters now
AI agents no longer wait for instructions. As designer Aurélie Radom points out, they go find their own context, references and first pass. The question used to be "what should this screen look like?" Now it is "which decisions do I keep for myself?"
Two wrong answers
"We don't use AI, it cheapens the work."
A slower, more expensive way to land on the same generic result everyone using AI already reached.
"We let the tool run and ship whatever it makes."
How you end up looking exactly like your competitor, built from the same defaults.
Both answers avoid the same question: where does a human belong in the process? Answer it on purpose, not by accident.
The decision matrix, stage by stage
- Concept. Automate mood boards, reference pulls and quick variations. Protect the point of view behind the concept.
- Prototype. Automate layout permutations, responsive resizing and placeholder content. Protect whether the interaction actually feels right, in the real product.
- Production. Automate asset generation, code scaffolding and first-draft copy. Protect brand consistency, accessibility tradeoffs and final sign-off.
- Launch. Automate A/B variants and performance reporting. Protect what ships and what gets pulled back, whatever early data says.
How to run this with your team
1. Audit last month
Pull one AI-made deliverable. Does it look like yours, or like everyone's?
2. Name the zones
Sort each step of your process into automate, protect, or decide case by case.
3. Write the rule down
Name three decisions a human must always make. Keep them visible before every project.
4. Run the critique loop
Before anything ships, score it against a fresh set of eyes that never saw the brief. Below a 9 out of 10, it goes back.
This article is based on The Human-in-the-Loop Design Guide to Claude 2026, our practical guide with the full decision matrix and the prompts design teams use with it.
This article reflects Engin Comakci's perspectives, prepared with AI writing assistance.
Setting the rules for AI in your product team?
Redesign helps teams decide what AI automates, what stays human, and how to move from AI pilots to everyday practice.