Automattic gave two-person teams 30 days to pause roadmap work, build something real, and ship it with full autonomy.
2
Sanja Grbic moved from design work into production code because she could own the full process and work directly in a prototype.
3
Engineers can increase their impact by teaching and enabling colleagues, while teams need access, experimentation time, and agency to change how they work.
Summary
Sanja Grbic describes Automattic's Radical Speed Month, a 30-day experiment in which about 500 employees started around 794 projects after pausing roadmap work. She built three products while learning to ship code at work. The first was a board game session manager built by a mixed group in two hours. It showed her how much an engineer can contribute by helping others learn. Her second project was a design system status tracker, which she built largely alone and deployed internally and publicly. Owning the whole process moved her from product designer toward design engineer. Her third project, an iOS chat proof of concept for WooCommerce merchants, went from idea to working product in six days with another designer. Across the projects, prototypes replaced high-fidelity Figma work as the starting point. Grbic argues that large organizations need access to tools, internal teachers, room to experiment, and enough agency for people to change established habits.
A short autonomous project can expose how much team structure slows delivery
Automattic paused roadmap work for 30 days and paired people to ship real projects. The first round included about 500 employees and started around 794 projects. Grbic says the experiment was possible within a fully distributed, asynchronous company with extensive documentation. Participants could use AI, but it was not required. The point was to create a temporary setting where small teams could make decisions, build software, and ship without the usual sequence of handoffs and negotiations.
Automattic prepared non-engineers to work with code before the experiment
Automattic first gave employees access to AI tools and courses, then introduced a two-week role-specific AI enablement course with lectures and hands-on work. As coding tools improved, everyone was encouraged to contribute to code. Systems and operations teams provided security processes, documentation, and a development setup that Grbic, a non-engineer, could use. She also relied on Context AC, an MCP server that gives AI tools access to the company's documented knowledge and data.
An engineer can have more impact by enabling teammates to contribute
Grbic's first project was a board game session manager built in two hours by two designers, an engineer, and a product lead. The app let people create, view, join, and manage sessions, see available games and spots, chat, and start or stop a session. The engineer set up GitHub, explained basic commands, and helped the group understand Git. Because the app was unrelated to an existing product, the group could focus on learning. Grbic's conclusion was that engineers need to become teachers and enablers when AI lets people work outside their usual domains.
Owning design and implementation can turn a designer into a design engineer
For her second project, Grbic worked with an engineer and a design operations colleague on ways to expose current design system information to people and AI tools. She proposed a tracker that collected links from GitHub, Storybook, and Figma, then sorted them by status and library name. After questions about performance and maintenance, she built a proof of concept herself in about two and a half weeks. The prototype took a week, followed by another week and a half for fixes, hosting, and deployment as an internal tool and a curated public site. She attributes the shift to AI and to owning the full process.
A working prototype can replace Figma as the first serious design artifact
Grbic's usual process began in Figma and stayed there until the design reached high fidelity. During Radical Speed Month, she and another designer planned an iOS chat for WooCommerce merchants, built the product, and returned to Figma for a mood board and visual adjustments. Their six-day proof of concept included WordPress.com authentication, a Jetpack connection, a site-themed visitor widget, and an AI agent that scanned site information and decided whether it could answer a question. Recording chats and ideas in the project folder also made collaboration and later building faster.
Shared tool knowledge lets small teams move faster together
The third project showed what happened after Grbic had already changed her process. She and another designer moved from zero through ideation and exploration to a working WooCommerce merchant chat in six days. Their alignment was relatively easy because both designers kept users in mind. They also had a shared understanding of how to apply the tools, so neither person had to wait for a separate design-to-engineering handoff. Grbic presents the speed as a result of both people learning and adjusting together.
Large organizations need human conditions for AI tools to change work
Grbic says tools alone will not change established processes in a large company. Organizations need to give people access to the tools, identify enablers and champions, create time and space for experiments, and give teams agency to make decisions. Her three projects suggest different effects: engineers can teach, designers can ship code in an established system, and pairs can develop a shared working method. She recommends starting at team level when company-wide process changes are difficult.
"The biggest insight, I think, for me from this project was that if you're an engineer, the impact that you have when you enable others may be far greater than the impact of doing more engineering yourself."08:36
Who should watch
You work in a large product organization where handoffs, approvals, or roadmap commitments limit small experiments.
You are a designer who wants to understand what changes when you build and deploy prototypes directly.
You lead engineers or mixed-skill teams and need practical ways to help colleagues use AI coding tools safely.
