AI in Practice for Development Teams

The training is meant for teams that already have the AI tools switched on and have found out that mere access to them does not raise productivity by itself. It focuses on what separates a demo from everyday practice: how to hand the tool the context, how to verify what it produced, and how to turn one-off enthusiasm into a habit of the whole team. Most of the time is spent working on code, ideally your own.

Goals of the Training

The goal is that after the training participants can choose a suitable type of AI tool for a given task on their own, prepare an environment in which an agent works reliably, and above all that they are able to verify and defend the result of its work. The course is built in modules — they can be assembled into a one-day introduction as well as into a longer tailored program.

Training content

Module 1. The landscape of AI tools: Chat, Assistant, Agent

Three levels of tooling that commonly hide behind the single word “AI”. What each of them can and cannot do, and where the boundary between them runs. This includes the difference between the model and the harness built around it — most of the confusion about “what AI can actually do” comes from exactly there.

Exercise: the same task with three tools – where the results diverge and why

Module 2. Agentic development in practice

Working with agents in the editor and on the command line (GitHub Copilot, Copilot CLI, Claude Code, Cursor). How to break a task down so that an agent can carry it, and when to stop it instead.

Exercise: implementing a new feature with an agent, from the assignment to the review

Module 3. Context is the product

Why the same agent performs completely differently on two codebases. Preparing the repository, documentation, architectural boundaries and instruction files so that the agent has something to work with.

Exercise: preparing a context file for your own repository

Module 4. Guardrails: tests, review and quality

The more trustworthy your automated checks are, the more autonomy you can safely hand over. Automated tests as the objective truth about what the agent produced; and the places where code review has to stay with a human.

Exercise: an agent on code without tests and with tests – comparing the result

Module 5. Security and data

What actually leaves the company the moment you send a prompt. How the enterprise variants of the tools differ and what to ask about them. And how to set rules that a team will realistically follow, instead of rules that get worked around.

Module 6. Making it stick

How to turn a one-off training into a habit: a community of practice, sharing prompts and context, and measuring impact without falling into the trap of counting tokens or accepted suggestions. This module is aimed at those who will be teaching the rest of the company.

Participant requirements

• practical software development experience (the course is not an introduction to programming)
• at least brief hands-on experience with some AI tool

Technical requirements (BYOD)

• your own laptop with the rights to install tools (with corporate devices please verify this in advance – it is the most common source of delay)
• access to at least one AI tool for development (GitHub Copilot, Claude Code, Cursor or similar)
• your own repository that can be worked on, if you want to practice on real code

The training is an introduction to the practice, not a replacement for it. It brings the best results in combination with follow-up technical coaching directly in the team, where the new habits settle on real work.