中文
Figma Config

Figma built an operating system for PMs in one month — the full, repeatable path from one person's write-up to 100% adoption

Skill issue: how we built an OS for PMs at Figma | Config 2026 · Janen

18 min
AgentContextAI Product

18 min total·Actually worth watching closely: ~4 min·2 must-watch clips

Orange = the 4 minutes worth watchingFor the rest, the guide is enough
Segment guide · 6 segments
  1. 0:12 3:20Listen

    The PM's fragmentation problem

    The opening lays out the pain: PMs switch context across five or six tools that don't talk to each other, their time eaten by chasing notifications and glue work like meeting prep, squeezing out the strategy and craft only a PM can do.

    The AI opportunity isn't replacing the PM — it's absorbing the glue work and giving the PM back to strategy and craft

    Pure narration, no key visuals — fine to listen to like a podcast▶ Jump to 0:12
  2. 3:20 7:17Listen

    The one-month challenge and the product principles

    How PMOS went from one PM's exploratory write-up to three volunteers building it, under clear CPO sponsorship and a deadline to roll it out to the entire team in one month, and the principles they set along the way: start with one use case, build for the happy path.

    Permission plus a time constraint breeds creativity; mind the split between having to use AI at work and getting to, and make sure AI adds value rather than being imposed

    Mostly narrative and principles with low visual dependency; the "happy path" framing around 5:17 is worth listening for▶ Jump to 3:20
  3. 7:18 10:25Watch

    The building-block architecture and the default skills

    Breaks down the core of PMOS: context, profile, and experience abstracted into reusable files that live together in a GitHub repo, plus demos of the three default starter skills — the "What's up" digest, the automatic weekly update, and the spec reviewer.

    Every prompt and skill draws on the same set of building blocks, and the combination is greater than the sum of its parts

    The best demo footage in the talk, especially at 9:40 where the system checks GitHub and the org chart to attribute contributions — listening only means missing the most convincing part▶ Jump to 7:18
  4. 10:25 14:18Skim

    Sitting with users, then hijacking an offsite

    The rollout method: be a forward-deployed engineer at your own company — sit with teammates, watch them onboard, keep a friction log and iterate. After adoption stalled at 25%, an offsite hackathon captured the team's full attention and took the product org to 100%.

    Breaking an internal tool's adoption ceiling isn't about more features — it's about earning your users' undivided attention

    Mostly slides and numbers you can skim, but stop at 13:44 for the 100% adoption moment and the story of Pratik rejoining the team▶ Jump to 10:25
  5. 14:18 16:25Listen

    In practice: preparing one meeting end to end

    A real product review as the worked example, chaining skills together: the Hex MCP for experiment results, the GitHub connector for architecture trade-offs, the Figma MCP for the FigJam agenda. The PM only defines the decision and how to frame the discussion — at least two hours saved each time.

    AI doesn't replace the discussion that matters; its value is making every piece of prep serve that discussion

    A narrated walkthrough with no strong visual dependency; focus on how the tools divide the work and hand off to each other▶ Jump to 14:18
  6. 16:25 17:36Listen

    Closing: no barrier, anyone can copy this

    The wrap-up: there is nothing unique about PMOS — the technology is nascent, there's no skill gap, and any team could build the same for itself as a PM OS, a design OS, or even a company OS.

    It's not too late to start: the method matters more than the tools, and this recipe travels to any team

    A closing argument with no visual dependency — listen through and take away the call to action▶ Jump to 16:25