Metamate
I designed the original Metamate end to end, then returned three years later to lead a broader reinvention of the product. The work moved from building Meta’s first internal AI assistant to aligning several organizations around a workspace for people, teams, and engineers.
The second chapter was as much an organizational design problem as a product design problem: the teams were not aligned, the product was still framed as chat, and the direction needed to become concrete enough to earn commitment and ship quickly.
Build an internal counterpart to ChatGPT
We started Metamate in 2023, a few months after ChatGPT went viral. Meta needed an internal counterpart, and I was asked to lead design for the effort. I worked as the sole designer and design lead, shaping the product from its first proposition through the shipped experience.
The system grew around three connected surfaces: a query-to-internal-search-to-answer flow, a broader platform, and an IDE plugin for engineers. Together they established the first useful product model for AI-assisted work inside Meta.
A product that became part of daily work
By mid-2024, Metamate had moved beyond experimentation. Approximately 96% of the company used it at least once a month, around 70% used it four or more times a week, and the coding experience reached roughly a 70% acceptance rate.
Those signals mattered because they changed the question from whether employees would use an internal AI product to how far the product could expand into their work.
Evolve the product beyond chat
When I returned to Metamate, multiple teams were coming together without a shared direction. The organization was not yet AI-native, and there was resistance to moving beyond the existing chat product.
I developed and socialized a longer-term direction, earning buy-in from general managers, vice presidents, directors, and teams across several organizations. The goal was to change the team’s frame from improving chat to building an extensible AI workspace.
Make the future tangible enough to fund
I pitched a product vision designed to excite leadership while giving the team a buildable path. Initial team buy-in was partial, so I split off a small crew to make the direction real rather than continuing to debate it in the abstract.
That crew built an iOS app, an internal software factory, and a new design system. The prototypes helped secure full leadership support—and created a four-week window to turn the vision into a shippable product.
Replace the original product
Metamate 2.0 shipped as the replacement for the original instance. The launch brought together a wider product system rather than another iteration of chat.
- Memory and Projects for individuals and teams
- A new harness, skill store, and design system
- A mobile app and expanded support for engineering work
We retired the previous product and increased retention. The response drove further investment: the design team grew by roughly five times, while product and engineering grew by roughly four times before the organization was restructured around the work.
Prototype earlier and protect the long term
In retrospect, I would spend more time prototyping before the final shipping push. The design team needed more protected time for long-term work, while the product itself needed a stronger bias toward simplifying complex flows as new capabilities arrived.
I would also treat prototypes as a continuing product artifact—updated over time as the direction, technical system, and organizational commitments changed.