Your team is learning AI from each other, not from you
Designers now learn AI mostly from peers, and leadership guidance has dropped as a source. That moves a design leader's job toward setting context and protecting experimentation.
Designer Fund's AI in Design 2026 report found that designers who learn AI from peers went from 24% to 80% year over year, while learning from leadership recommendations fell from 32% to 16%. Nobody on a design team is waiting for the leader to hand down the right workflow, and the ones with the best setups are the ones sitting next to you, so the leader's job is to make sure what one person figures out on Tuesday is usable by the whole team on Wednesday.
The same report found that the average designer now uses 7 AI tools, up from 3. Seven tools per person across a team of eight is dozens of private setups, and most of what makes each one work lives in a single head.
Write the context down where the next session starts
The cheapest way to share a workflow is to write down the context an agent needs and put it where every session starts, whether that is a file in the repo or a shared doc the tool loads automatically. Take your best prompt setup and write down what it assumes: which components to use, which colors are tokens, what a finished state looks like, what must never be hand-rolled. Then watch for the second correction. When a designer fixes the same agent mistake twice, the fix belongs in the shared instructions, because a fix in the file reaches every session that loads it and a fix in someone's head reaches the one session they are running.
Make sharing a habit
Learning from peers happens when it is cheap to do and visible when someone does it. A standing slot in the team meeting for one workflow, shown on real work and kept to five minutes, will do more than a training program. The format that works: what I was trying to do, what I gave the agent, what came back, what I changed. Ask people to bring a failure along with a win, since a failed attempt teaches the room where the edges are. Keep the prompt, the instruction file, and the before and after in one searchable place, so a person who missed the meeting can still pick the workflow up. I find out where work is drifting through shared docs, dashboards, and standups, and a workflow that lives in a shared doc spreads for the same reason: people see it.
Credit the teachers
If sharing a workflow takes an hour of unpaid effort and earns nothing, people will stop doing it. Name the person whose setup the team adopted, say what it saved, and carry it into calibration as leadership. The senior row on your ladder should mention it, and so should the promotion packet. I advocate for my reports in calibration, and evidence that several teammates now work the way one person's setup works is the kind that wins that room.
Let people break things
A team that learns from itself makes mistakes in public, and the leader sets how that goes. I treat mistakes as a sign of moving fast, and the only thing I discipline is unethical behavior. A designer who shares an experiment that went sideways and gets thanks has told the next person it is safe to bring theirs, and a designer who gets a postmortem has told everyone to go back to private setups.
Be the first to rate this article.
Let's work together.
Open to select projects and collaborations — design systems, accessibility, and AI-native product work.