Playbook series: Communities of practice

Published via GitHub Executive Insights | Authored by Matt Nigh, Program Manager Director of AI for Everyone

In this series reflecting on GitHub’s internal playbook for AI adoption we have covered how to establish clear guardrails, provide learning opportunities, and activate a network of advocates. Now we’re digging into how you scale day-to-day learning to ensure valuable knowledge doesn’t vanish into private chats.

The solution lies in Communities of Practice (CoPs). If your AI Advocates are the catalysts, CoPs are the chemical reaction. They are the forums where individual discoveries become collective intelligence, and where your company's AI fluency truly grows.

This guide shows you how to move from scattered, one-off conversations to intentional, high-impact AI communities.

The Problem: Fragmented and lost learnings

AI experiments are already happening inside DMs, in private team Slack channels, and buried in slide decks. But most of that insight disappears fast. A clever prompt is forgotten. A smart workaround is shared once and never again. Someone solves a problem, but the next team has to rebuild from scratch.

Without a system to surface and spread these moments, you get duplication instead of acceleration. A solid CoP strategy fixes that. It gives your organization a shared, searchable, and social space to learn together.

A blueprint for your communities

Cultivating a CoP network is an act of intentional community management. It’s about providing the right structure and then actively nurturing the conversations within it.

Step 1: Start with a hub-and-spoke model

Don't start with a single, monolithic #ai channel. It will quickly become too noisy to be useful. Instead, launch with a "hub-and-spoke" model to serve different needs.

• The hub: A general-purpose entry point. Create a single, company-wide channel (e.g., #how-do-i-ai) to serve as the front door for your entire AI ecosystem. This is the place for broad announcements, general questions, and for employees who don't know where else to go.
• The spokes: These should be deep-dive, specialist communities. Launch one or two specialized channels for your most critical user groups. A developer-focused community (e.g., #copilot-power-users) is almost always a good place to start. As adoption grows, you can spin up new "spokes" for other functions like marketing, sales, or finance, as demand grows.

Step 2: Assign clear ownership and a simple charter

Every community needs a designated leader and a clear purpose. This is a perfect role for your AI Advocates.

• Assign a leader: Ask an Advocate from the relevant department to be the official owner or facilitator of each community.
• Write a charter: The leader should write a simple, one-paragraph charter pinned to the channel. It should answer three questions: What is the purpose of this community? Who is it for? What are the key topics of discussion? This prevents confusion and keeps the conversation focused.

Step 3: Actively manage and sustain momentum

Creating channels is easy, keeping them valuable is hard. The AI Program DRI, in partnership with the community leaders, must actively work to sustain momentum.

• Showcase and amplify: Create a rhythm for finding and celebrating the best content from your CoPs. This could be a weekly "best of" post that summarizes the most interesting use cases or a "discussion firehose" that automatically shares tagged posts from various repos into a central channel.
• Run community programming: Use the communities as the primary venue for your AI events. Announce new L&D opportunities, host tool-specific Q&As, and run your office hours within the relevant channels to drive engagement.
• Facilitate connections: The DRI and Advocates should actively monitor conversations to act as connectors. When one person asks a question, they can tag another employee who they know has the answer. This active facilitation builds the network and reinforces the value of the community.
• Encourage leadership participation: Momentum is not just a grassroots effort; it needs air cover. Leaders and senior ICs play a critical role in sustaining community health. They should be encouraged not only to direct their teams to these communities but to participate themselves. A senior engineer asking a question in the developer channel or a VP celebrating a use case shared in the general hub provides powerful social proof, signaling that these communities are a valuable and important place to spend time.

Step 4: Supercharge your communities with events

Async channels are the foundation, but live and structured events are what turn a community into a vibrant hub of innovation. These events create focal points for learning and strengthen the connections between members.

• Host virtual learning events: Organize company-wide events like a "Day of Learning" or a hackathon focused on solving specific business problems with AI. These events, run through the CoPs, generate a burst of energy, collaboration, and a wealth of new use cases. Crucially, all sessions should be recorded and shared for asynchronous consumption, turning a one-time event into a lasting learning artifact.
• Run executive listening sessions: Create a direct feedback loop by hosting regular, informal listening sessions between community leaders and executive sponsors. This gives leaders an unfiltered view of what's working, what's not, and how they can best support the company's AI and L&D efforts. It also provides community leaders with a direct line to decision-makers, making them feel valued and empowered.
• Organize community leader offsites: Bring your community leaders (your AI Advocates) together for a dedicated virtual offsite or summit. This provides a valuable opportunity for them to share best practices, align on goals for the next quarter, and build the personal relationships that are essential for a strong, self-sustaining network.

When you shift from a passive approach to actively nurturing community, those scattered conversations start to work together. You build a space where people learn from each other, knowledge flows more freely, and AI fluency spreads across your organization.

Complementing CoPs with tooling

Your CoPs generate the knowledge, and these tools ensure it’s captured, refined, and shared effectively. Each tool play a distinct role in how knowledge is created, shared, and scaled across your organization.

• Slack/Teams channels are your CoP’s real-time pulse, where ideas come alive, problems are solved, and questions find quick answers.
• GitHub Discussions (or similar) provide structure and continuity—a searchable home for insights you don’t want getting buried in chat.
• Internal wikis are where best practices, reusable assets, and canonical guidance live on.

Here’s how to connect them:

• Capture valuable insights: When a conversation in chat surfaces a breakthrough, move it to GitHub Discussions. Add a clear title, summary, and tags like prompt, copilot-tip, or use-case which will also make content easy to discover later.
• Foster async participation: GitHub Discussions support thoughtful input across time zones. They’re ideal for deeper problem-solving and design debates.
• Automate highlight sharing: Use lightweight bots to summarize top GitHub Discussions on a regular basis or auto-post starred threads into CoP channels.
• Build cross-functional bridges: Encourage cross-posting between functions, like surfacing a product team’s AI win in an engineering CoP, or vice versa.
• Promote graduation to wiki: Once a discussion gains consensus or maturity, document it in your internal wiki. Include links to the original thread for context.

Make learning visible, scalable, and durable

If you want AI adoption to stick, you need more than tools and training. You need a system that captures what people learn, connects those learnings across teams, and turns them into reusable knowledge.

Communities of Practice give you that system. They make day-to-day learning visible. They prevent repetition by sharing solutions and accelerate progress by connecting people solving similar problems. And when combined with structured tools like GitHub Discussions and internal wikis, they ensure your best ideas don’t get lost.

Don’t let AI learning stay locked in private chats or isolated teams. Build a network where insights are shared, refined, and documented. That’s how you move from scattered experimentation to coordinated transformation.

AI fluency at scale doesn’t happen by accident. It happens when you design for knowledge flow, sustain community energy, and make collective learning part of how your company works.