Define what participants will be able to do by the end
Before choosing a tool or a slide deck, write down an observable action: run a test, evaluate a RAG answer, change a component or deploy an artifact. “Get to know AI” does not make clear what the team needs to be able to do. A defined outcome helps you choose the right exercise and depth.
Carnegie Mellon's guidance on instructional design emphasizes alignment between objectives, activities and assessment. For a development team workshop, that means the exercise should let participants apply the skill, and the wrap-up should test that same skill. An instructor's demo on its own does not show that the team can carry on without them.
Check the starting point and the work environment
Learn the participants' level, their roles and the tools they are allowed to use. An exercise that assumes Docker experience or access to every cloud service may not fit. Set prerequisites and run an environment check before the workshop.
Use synthetic data and approved accounts. There is no need to upload client documents into a new tool to practice working with AI. If the organization's environment restricts installations or network access, prepare a suitable path and do not count on bypassing restrictions during the session.
Build an exercise that involves a decision and feedback
A hands-on exercise needs a task, constraints and a way to check the result. For example: fix a failing API call, add a permission check and describe what changed. Using a coding agent is part of the process, and participants still need to read the change and demonstrate the behavior.
Include an example that does not work the first time. It creates an opportunity to practice diagnosis, reading errors and revisiting assumptions. Make sure the difficulty serves the learning objective; an installation problem unrelated to the topic can eat up time without developing the skill you were aiming for.
Assess independence, not just following instructions
After the practice, change a small condition and ask participants to adapt their solution. Someone who can rerun the same command does not necessarily understand why it fits. Ask for a brief explanation of a choice, a limitation or a test they added.
Feedback can take the form of a discussion, an output review or solving problems in teams. Not every workshop needs to become a formal exam. What matters is knowing which parts participants can do on their own and where they need more practice, so you can plan next steps on solid ground.
Plan what remains after the session
Hand over run instructions, practice materials, tests and safe examples people can come back to. Define a small follow-up task that connects to real work, and an owner who can help if someone gets stuck. Without follow-up, the output may end up living only in the workshop folder.
- A clear, practical learning objective for the team.
- An environment that can be run without exposing sensitive information.
- An exercise in which a decision is made and the result is verified.
- Follow-up materials and a chance to practice independently.