app-entwickeln-lassen.de LogoApp entwickeln lassen

Workshops & Training

Custom AI Workshop on Your Own Codebase: Learning with Real Projects

app-entwickeln-lassen.de editorial team 6 min read
Contents
  1. Why Sample Projects Often Fall Short
  2. Sample Project or Your Own Code: A Comparison
  3. What Actually Happens in a Workshop on Your Code
  4. Prerequisites: Data Protection, Approvals and Confidentiality
  5. Choosing Suitable Tasks
  6. What Remains After the Workshop
  7. Organisation and Cost
  8. Conclusion
  9. Frequently Asked Questions

An AI workshop on your own codebase means your team uses coding agents not on generic sample projects but directly on your repository, your conventions and your real tickets. The benefit: what is learned transfers straight into daily work, and the training ends with concrete results in the code. This article explains how such a workshop works, what you need to settle beforehand and where its limits lie.

Why Sample Projects Often Fall Short

Many AI training courses use clean demo repositories: a small codebase, good tests, modern libraries. Coding agents perform impressively on them. Reality in companies looks different. Grown codebases have their own conventions, historical baggage, internal libraries and gaps in test coverage. That is exactly where it becomes clear whether an agent helps or creates chaos.

Teams that have only practised on sample projects are often disappointed when they turn to their own repository: the agent ignores conventions, cannot find the right files or proposes solutions that clash with the existing architecture. The real skills – building context, slicing tasks, reviewing results – are best learned where they will be needed later.

Sample Project or Your Own Code: A Comparison

AspectGeneric sample projectYour own code
PreparationMinimal, ready to goHigher: approvals, access, task selection
RealismLimited, "fair-weather conditions"High, including legacy and quirks
Transfer to daily workLeft to the teamDirect, results stay in the repository
Data protection effortHardly relevantMust be settled in advance
Best forFirst exposure, mixed groupsTeams ready to switch in earnest

The two approaches combine well: a short start on a sample project covers the basics, then the team moves into its own codebase.

What Actually Happens in a Workshop on Your Code

A custom workshop does not follow a rigid curriculum; it is shaped by your repository and your goals. Typical steps:

  1. Let the agent explore the codebase: The team asks an agent to explain the architecture and checks where it is right and where it is wrong. This immediately shows which context is missing.
  2. Create context files: Together, the team drafts first project rules with build commands, conventions and architecture notes.
  3. Deliver real tickets: Small features, bug fixes or missing tests from the backlog are tackled with agents.
  4. Review results: Every change is reviewed together, which sharpens the eye for typical flaws in agent-generated code.
  5. Document limits: Where did the agent fail and why? This produces rules for daily work.

Tasks from older parts of the codebase are especially instructive. If modernisation is your focus, see agentic coding workshops for legacy modernisation.

Prerequisites: Data Protection, Approvals and Confidentiality

When an external trainer and a coding agent work with your code, a few questions must be settled beforehand:

  • Confidentiality: A non-disclosure agreement with the trainer is common and sensible.
  • Tool approval: Which coding agents may access your code, and how does the provider process the data?
  • Personal data: If the repository contains test data, logs or configuration with personal data, remove or replace it first.
  • Secrets: Credentials do not belong in the repository – and certainly not in a workshop. Check this beforehand.
  • Scope: Does it have to be the entire repository, or is a bounded module enough?

How to decide whether source code may be sent to cloud services is explained in AI coding tools and data protection. This is not legal advice; involve your data protection officer if in doubt. A detailed checklist for organisational preparation is available in agentic coding workshop preparation.

Choosing Suitable Tasks

Not every ticket suits a workshop. Good candidates have a clear definition of done, can be solved in manageable time and can be verified by tests or a brief manual check. Proven examples:

  • Unit tests for a previously untested module
  • A small, well-described feature
  • A reproducible bug with a clear error message
  • A bounded refactoring, such as extracting a component
  • Updating a dependency including adapting the call sites

Large, vague tasks or changes to security-critical areas without test coverage are less suitable. They work better as a discussion of how to prepare them for agents.

What Remains After the Workshop

The biggest benefit of a workshop on your own code: it leaves artefacts the team can keep working with immediately.

  • Context files and project rules that live in the repository and continue to be maintained
  • Several delivered and reviewed tickets
  • A review checklist tailored to your codebase
  • A list of areas where agents work well and areas that need preparation first
  • Concrete next steps, such as additional tests in critical modules

How to develop context files further afterwards is described in project rules and context files for coding agents.

Organisation and Cost

At app-entwickeln-lassen.de we place AI developers from real-world practice as trainers who can familiarise themselves with your codebase in advance. Content, format (on-site or remote) and scope are agreed individually. Workshops start from €799 net; because a workshop on your own code needs more preparation, the price depends on the effort involved. You receive a quote on request by e-mail.

Planning an app? Request a free fixed-price quote in 2 minutes

Conclusion

An AI workshop on your own codebase takes more preparation than standard training but delivers much more: realistic experience, direct transfer and concrete results in the repository. Settle data protection, approvals and confidentiality carefully, pick suitable tickets and plan how the rules you create will be maintained afterwards.

Frequently Asked Questions

Does the trainer need to see our code before the workshop?

It is recommended so exercises can be prepared properly. Access should be covered by a non-disclosure agreement and limited to what is necessary.

What if our code is in poor shape?

Then a workshop on your own code is especially instructive. The team learns how to use agents deliberately in difficult codebases and which groundwork, such as tests, pays off.

Can a workshop on our own code be run remotely?

Yes. All participants need access to the repository and the approved tools, and screen sharing must work reliably.

Is code produced in the workshop merged directly?

Only after regular review. Workshop results should go through the same quality process as any other code.

This article was written by the editorial team of app-entwickeln-lassen.de – the app and AI development service of jocapps GmbH, based in Freiburg, Germany, and building apps since 2014. The content is for general information and does not replace individual (legal) advice.

You might also like

Workshops & TrainingAfter the Workshop: Embedding Agentic Coding in Daily Work

Embedding agentic coding after the workshop: owners, team rules, rituals and metrics that turn a one-off training into a lasting way of working.

5 min read
Workshops & TrainingAI Coding Workshop for Startups: Reach Your MVP Faster with Agents

AI coding workshop for startups: how small teams use coding agents to reach their MVP faster without piling up technical debt – content, formats and cost.

5 min read
Workshops & TrainingAgentic Coding Training for the Public Sector and Its IT Providers

Agentic coding training for the public sector and its IT providers: data sovereignty, AI literacy under the AI Act and secure formats – explained in practice.

6 min read
Workshops & TrainingHow to Run an AI Hackathon with Coding Agents: A Practical Guide

How to run an AI hackathon with coding agents: goals, challenges, security rules and judging – so the event becomes a real starting point for agentic coding.

6 min read
Workshops & TrainingRemote Agentic Coding Workshops: Success Factors for Distributed Teams

Remote agentic coding workshop: tech setup, session flow, interaction and follow-up – how distributed teams learn coding agents online as well as on-site.

5 min read
Workshops & TrainingAgentic Coding Bootcamp: Intensive Training for Developer Teams

Agentic coding bootcamp for developer teams: when intensive training pays off, how to structure it and what your team can actually do afterwards.

7 min read