Why Create an AI Community of Practice?
Many people are learning to use generative AI tools as they go. Whether they’re curious, enthusiastic or concerned, they’re experimenting, developing their own methods and beginning to figure out how AI fits into their work. These people are generally known as early adopters.
During this transition, I keep coming back to one question: how can we do a better job of sharing what we learn so that everyone doesn’t have to keep starting from scratch?
An AI community of practice brings together people who experiment and regularly share their methods, results and challenges. It can help team members learn from one another, provided they share enough context for others to put those lessons into practice.
I’ve been exploring AI’s possibilities for quite some time. I’ve tried various tools and conducted a range of experiments, including a research project where I teach. Along the way, I’ve discovered promising uses, but also limitations and mistakes I wish I’d understood sooner.
Knowledge That’s Still Difficult to Share
In the settings where I work, I see people developing their own ways of using AI. They refine their prompts, run more tests and eventually find an approach that works. Much of that knowledge, however, stays in their heads, in a chat history or in files that are difficult to find.
Someone elsewhere may then spend considerable time trying to solve the same problem, unaware that a colleague, an instructor or someone else has already explored it.
What’s missing is a way to share the process behind the result. Knowing which tool someone used isn’t enough to reproduce their work. You need to understand the task, the tests they ran, how they checked the output and the limitations of the final result.
A Conversation That Broadened My View of AI Communities of Practice
I had a memorable conversation with someone in the Canadian Armed Forces who shares my enthusiasm for exploring AI. We’d carried out several similar experiments, but approached them differently. My approach was more technical; theirs was more artistic.
We compared our methods and discussed what had worked and what hadn’t. Those different perspectives made the exchange particularly valuable. Each of us brought something the other might not have discovered alone. That conversation made me want to have more exchanges about it. People experimenting in a college, a small business or a community organization sometimes face similar challenges, even when their day-to-day work seems worlds apart.
An AI Community of Practice Grounded in Everyday Work
For me, a community of practice should give people a chance to share actual work and talk it through. Someone could present a task, explain why they used AI and show the corrections they made.
There’s also value in sharing experiments that were abandoned. A convincing but inaccurate response can teach us something about the checks we need to build into our work. An unreliable automation can help us recognize situations where human involvement is still the better choice.
Initiatives like this already exist in Québec. Mila’s Community of Practice, for example, brings people from research and industry together to discuss the challenges of applied AI. The format can vary depending on who’s involved. My main question is how to make these exchanges accessible in the settings where I work.
For people who want to become more comfortable with AI and use it in their everyday lives, La Cuisine d’Info IA Québec, where I contribute as a trainer, offers a learning community with short lessons and opportunities for members to exchange ideas. People can learn at their own pace, ask questions and share their experiences with practical uses of AI.
Setting Rules for AI Also Means Showing People How to Use It Well
The guidance documents I read tend to focus on restrictions. They prohibit certain practices, remind people not to share confidential information, refer to Québec’s Law 25 and specify which tools are approved.
These rules are necessary to protect personal information and prevent irresponsible use. But they aren’t enough to help people use AI effectively in their work.
We explain what people shouldn’t do, but less often how to use AI well. I still see too little guidance on effective working methods, concrete use cases or responsible ways to incorporate AI into business processes, for example. There’s extensive discussion of risks and restrictions, but much less about how these tools can improve the quality of our work.
An AI community of practice can help meet that need by bringing together documented experiences and turning individual experiments into practical guidance for others.
Documenting an Initial Test So Others Can Repeat It
I suggest starting with a short record of the experiment. Each person could note what others need to know to understand the test and assess whether the approach would be useful in a different context.
Take a hypothetical example: someone uses an AI assistant to summarize a meeting. Sharing the prompt alone leaves several questions unanswered.
🤔 Does the summary correctly identify who made each decision?
🤔 Does it distinguish a proposal from a commitment?
🤔 Have the responsibilities and deadlines been checked against the original notes?
Documenting these points allows another team to assess the method before adopting it. If the goal is to save time, preparation and review time also need to be factored in. Something generated in seconds may require extensive corrections.
Starting with a Group and a Specific Task
I don’t yet know which format works best in every setting. I would start small, though: a few people, a shared problem and a scheduled time to compare their tests.
A first meeting could focus on a single task, such as designing a training session or preparing a working document, with each participant explaining their approach. The group could then identify what to test next, which precautions to take and which questions remain unanswered.
It would help to designate someone to keep a record of the discussion and arrange a follow-up after the tests. Without that follow-up, people may leave with interesting ideas that never get put into practice. A well-organized shared folder can be enough to get started. Decide how the group will work before choosing a platform.
Sharing What We Learn Across Different Settings
I’ve also worked with a regional soccer association in the Québec City area that is interested in establishing guidelines for AI use and has offered training. That experience raises the same question for me: how could what one association learns benefit others?
A method will sometimes need to be adapted. Needs, resources and obligations vary from one organization to another. Sharing AI knowledge becomes useful when those differences are explained and people can judge what applies to their own circumstances.
As an instructor, I want my experiments to benefit my students, my colleagues and people who are just beginning to explore these tools. Training and AI literacy have a role to play because they give people the knowledge they need to understand the methods being shared and assess them critically.
Creating an AI community of practice is one way to continue this work together. I would start with a simple question: what recent experience could you document to save someone else from starting from scratch?
Would you like to develop a more structured approach to AI within your team and help others learn from your working methods? Visit my website to learn about my training and coaching services, or email me at info@jimmygilbert.com to discuss your needs.
Jimmy Gilbert, Technology Consultant and Trainer. A computer science instructor at Cégep de Sainte-Foy, Jimmy brings together expertise in teaching, software development and artificial intelligence. He helps prepare the next generation of technology professionals by combining technical rigour with a practical understanding of industry needs, while actively incorporating AI tools into his teaching. He also works as a technology coach, helping individuals and professionals make effective use of technology, automation and AI.
