Since Part I of the Multiplayer AI series, our HR heroine, Priya, has really doubled down and become more fluent with AI for her own use cases.
She now has multiple routines running on Claude and gets excited about building automations, like pre-populating all sorts of data for annual compensation reviews to make the process faster and less painful.
Now, to fulfill her company's AI mandate, she has to think bigger.
The next challenge is figuring out how she and other early adopters of AI can stop experimenting in isolation, start collaborating to reduce re-work, and deploy skills and workflows in ways that will benefit everyone.
Here's what's happened so far:
A lot of people are posting about how they're using AI in a company Slack channel.
In fact, multiple people are posting about building a Skill to solve the same problem.
Everyone is manually installing a shared Skill or Plug-In from zip files, so there are 19 different versions floating out there.
People are doing a lot of great work, but that work isn't universally visible or up-to-date.
In gaming terms: People are playing on single-player mode.
Themselves against the game. Level up. Save time. Be more productive.
Priya needs to help them rebuild around multiplayer mode.
Multiple people show up, start with a tutorial to know how to use the basics, have agreed upon norms or rules, share resources, and battle together to achieve a shared objective.
​
(Can anyone tell I've been listening to a lot of Dungeon Crawler Carl lately?)
Why “Just Show Them What You Did” Doesn’t Work
Sharing wins is step one of most adoption playbooks.
But a screenshot of a great output teaches the same lesson as a magic trick: “this is possible,” not “here’s how you’d do it.”
It creates spectators, not players.
Spectators stay tourists. (Read Part I if the four-level AI ladder — Tourist, Resident, Citizen, City Planner — doesn’t ring a bell.)
The good news is that turning individual work into shared work isn’t a culture problem you have to will into existence.
It’s an infrastructure problem, and the infrastructure already exists.
The Sharing Stack
- Organizational Skills. With Claude, an Organizational Skill is the smallest reusable unit. It's a packaged set of instructions for a specific job (“draft a comp letter,” “summarize this week’s exit interviews”) that anyone on the team can trigger by name, not just the person who wrote it. To upgrade something to an Organizational Skill, you want it to be relevant to a critical mass of people and to have a specific author who is responsible for making revisions.
- GitHub. Skills, context files, and CLAUDE.md setups are just text, which means that just like GoogleDocs, they benefit from version control to show you who edited what when. If you use a Github repository (also called a "repo"), you can see a change log and have a review process (someone who looks at and approves a pull request or PR before it's merged and updated). If something breaks, you can also fix it once and have that fix sync to everyone. You can read more about how non-engineers are using Github here.
- The Marketplace. This is a distribution layer, or a directory your team can browse and install from instead of hunting through DMs with a colleague only to eventually say “hey, can you send me that prompt again.” Skills and MCP connectors get bundled into plugins, and plugins get grouped into the marketplace (either internal to your company or public). For your internal marketplace, you have sharing controls (e.g., auto-installed, someone can see it and choose to install, only some people can install it, etc.).
Put those three together and Priya’s company-wide AI Slack channel becomes closer to an app store.
If you're on a Claude Enterprise or Team Plan, you can also get detailed analytics on the # of uses of each Skill, the cost per use, etc. to figure out where there's demand for a specific type of workflow and whether or not that specific workflow has been cost-optimized by an expert.
The Four-T Playbook
Every edition, I share a proven tip, trick, tactic, or template. This time, it's a:
👉 Tip: Create a shared repo today — something like people-ai-skills — even before you have five things to put in it.
​
Sometimes, it's easier to start with the end in mind. This'll save you from having to migrate skills and workflows from 5 local environments, 3 Notion pages, and who even knows how many Slack channels.
Final thoughts
In Part III of The Multiplayer AI Playbook, we'll explore how businesses run day-to-day is changing with AI.
Until next time,
Melissa
P.S. This newsletter represents opinions and experiences from various HR professionals, it is not advice. Please always consult with appropriate legal counsel on AI use at your company.