The productivity gains from AI tend to concentrate in the hands of a few power users while less fluent product builders fall behind. Closing that gap takes more than tool access, and more than individual practice: teams at Figma, Sephora, Mercari, and Atlassian have each built a recurring forum where beginners and skilled practitioners learn together in a low-pressure setting.
Figma: peer-led study groups
Jake Geller, a design operations manager at Figma, piloted a study group called “AI Collective” after observing wide variance in technical and AI fluency across the design team. He surveyed designers on the skills they wanted to build and the learning environment they preferred. Interests ranged from becoming fluent in a single tool, such as Figma Make or Figma Weave, to shipping code and building custom plugins, but respondents agreed on one thing: small, psychologically safe groups over large team trainings.
Geller split designers into five groups, each organized around a different skill.
- Goal: Raise the baseline for less AI-fluent designers through hands-on practice and peer support.
- Format: Five peer-led groups built from the topics designers chose themselves. Leads volunteered, and participants picked their group. Each met for one hour a week over six weeks, with Slack keeping the conversation going between sessions. A flexible planning template suggested 20 minutes of teaching and 40 minutes of practice.
- Takeaway: A concrete goal, such as building a Figma plugin, gives participants a specific outcome to work toward and makes peer support the natural route to getting unstuck.
To replicate it, ask peers which AI tools interest them most, recruit volunteers fluent in those tools, and have each leader align the group around a tangible result like a finished prototype.
Sephora: show the process, not the polish
Sneha Narahalli, Sephora’s senior vice president of technology, runs a biweekly forum where builders across the organization present AI work, whether professional or personal. Early sessions featured only finished results—“the final glamorous AI-generated outcome,” as Narahalli puts it—and attendees reported feeling overwhelmed by polished peer projects without seeing a path to their own. The format changed so that presenters now walk through the process behind the work, including failed attempts. “Learnings aren’t just the good stuff,” she says.
- Goal: Make AI experimentation feel approachable by exposing the process, not just the output.
- Format: Every two weeks, participants share something they are building—an app made with a vibe coding tool, or skills that automate routine tasks—and cover what they tried and what didn’t work. Dedicated Slack and Teams channels sustain the discussion between meetings.
- Takeaway: Step-by-step process sharing gives newcomers a jumping-off point, and peer voices carry more weight than top-down messaging.
The low-lift version: reserve 10 minutes of an existing meeting, spin up a separate session, or open a Slack channel. Figma runs this as a “How I AI” segment in its bi-weekly all-hands alongside a channel of the same name.
Mercari: cross-training designers and developers
According to Aki Alkhatib, a principal designer at Mercari, AI makes building easier, but people don’t automatically know how to judge what they’ve made—especially when they’re working outside their own discipline. Mercari’s answer is two-way training.
- Goal: Help builders evaluate work produced outside their domain and collaborate more efficiently across functions.
- Format: Hands-on sessions in which designers learn engineering fundamentals—writing code, using an AI coding tool, setting up a GitHub repository—while developers study design basics such as visual hierarchy.
- Takeaway: Cross-training gives people enough context to carry an idea further on their own. A developer might draft an interface in Figma Make and apply visual hierarchy principles before handing it to a designer, who then applies deeper expertise to review and refine. “The starting point could come from anyone,” Alkhatib says.
Start with one basic design task for developers and one basic engineering task for designers, run a few expert-led sessions, and review the results as a group.
Atlassian: prototype first, explain second
Charlie Sutton, chief design officer at Atlassian, noticed that as teams adopted AI prototyping tools, designers typically explained context and rationale before revealing a prototype—if the team saw one at all. His fix is a simple rule: the prototype comes first.
- Goal: Build familiarity with prototypes, and their value, through regular practice.
- Format: A designer opens each crit by letting the group click through a prototype, which might be built in a tool like Figma Make. Only afterward does the presenter cover goals, research, and rationale. The rule doubles as routine practice with an AI tool, since every crit requires a prototype to show.
- Takeaway: As Sutton frames it: “Show me the thing, help me play with it as a prototype, and then explain to me why you made that decision.” Simple rules, he notes, are what make complex changes possible—though the team still needs periodic reminders to keep the habit reinforced.
To test it, apply the prototype-first rule to a single recurring critique for a month, let the group click through before any context is shared, and keep reinforcing until it sticks.
What these rituals share is a shift from individual experimentation to shared practice. Access to tools is the starting condition; recurring, peer-driven forums are what turn AI into part of how a team’s work gets done.



