Training · In-house
AI training for development teams. On your own code.
We maximise the productivity of your developers with the approach we use ourselves to build business software: agents implement in parallel, agents verify, your team decides. In one to two days, on a sample repository or directly in your code.
- 3–12 people
- 1–2 days
- In-house
- With certificate
Does your team use Claude Code? To the Claude Code training
01 — Starting point
The licences are bought. The way of working is still missing.
Copilot, Claude Code or Cursor are rolled out quickly. What happens afterwards looks similar in almost every intro call.
01
Everyone works differently
One person writes whole features with the agent, another only uses autocomplete. Prompts sit in private notes, results are not reproducible, and nobody can say what counts as good practice in the team.
02
Quality depends on chance
Pull requests get larger, reviews do not get more thorough. Whether generated code meets your own standards depends on who reads it – and how much time they have at that moment.
03
Security and privacy are unresolved
What may the agent execute? Which data leaves the company? Which MCP server from the internet is currently running with access to the repository? As long as this is unresolved, usage remains a grey area.
This is neither a tooling problem nor a motivation problem. What is missing are shared rules for what an agent may do, how its work is verified and who stands behind it in the end.
02 — What the workshop is about
Your team keeps architecture, quality and responsibility. The AI implements.
AI-assisted programming does not mean handing over decisions. It means making them earlier and more clearly: in rules, specifications and verification steps that every agent must follow. The four goals of the workshop are based on this.
Safe use of AI
Clear rules for what agents may do. Your team stays in control.
High-quality code
AI-generated code that meets your standards: clean, tested, maintainable.
Effective verification
Tests, reviews and quality gates that secure every change.
Data protection from the start
Software that meets data protection requirements, even with AI in the process.
03 — The new role
From writing code to steering. How the daily work of your developers changes.
This is exactly the shift we practise in the workshop: describing tasks so that agents can complete them, and building verification steps so that only tested work comes back.
- Writes the code
- Tests by hand
- Waits for the review
- Reworks
Each step waits for the previous one.
Build agents
- Interface
- API
- Data model
QA agents
- Tests
- Code review
- Security
Agents work at the same time, the developer reviews the result.
04 — What your team learns
We maximise the productivity of your developers. With four approaches from our projects.
A team gets faster through automation and parallelisation. It stays faster only if quality and control grow with it. That is why all four belong together.
01Automation
Recurring work runs by itself.
Creating a migration, writing a test case, preparing a review: described once as a skill or subagent, then used by the whole team.
Afterwards in the repository: Skills and subagents for your standard tasks
02Parallelisation
Several agents, one ticket.
Interface, API and tests are built at the same time instead of one after another. Your team learns to cut tasks so that agents can complete them in parallel.
Afterwards in the repository: Task breakdown and orchestration
03Quality
High quality, even though agents write the code.
Specification first, quality gates in the pipeline and an independent review agent. Whether a human or an agent wrote the code makes no difference to the gate.
Afterwards in the repository: Spec template, quality gates, review steps
04Control
Rules instead of chance.
Architecture and conventions in an AGENTS.md, clear permissions for agents and MCP servers, privacy clarified. Responsibility stays with your team.
Afterwards in the repository: AGENTS.md, permissions, vetted MCP integration
Your stack, your tools
We practise on a sample repository or on your own code. Previous workshops ran with Java and PHP teams, among others. The focus is Claude Code; the approaches apply to all common tools:
- Claude Code
- GitHub Copilot
- Cursor
- Codex
05 — Key facts
AI workshop for your development team: the framework at a glance.
- Location
- In-house
- On site at your company
- Group size
- 3–12 people
- From 6 people with two trainers
- Duration
- 1–2 days
- Depending on topics, recommended: 2 days
- Preparation
- Briefing call
- Topics matched to your team
- Format
- Hands-on
- Theory, applied in practical exercises
- Completion
- Certificate
- For all participants
We usually reply on the same working day.
06 — Why Viewconic
We train what we use ourselves. Every day, in client projects.
We have been using this approach ourselves for some time, in production client projects. The figures:
- production client projects, built purely with AI
- 15
- tickets completed entirely with AI
- 658
- at a Munich patent administrator
- 9instead of 75 days
- of the effort estimated the classic way, our experience
- ⅓–⅕
As of: November 2026
The workshops are led by David Kreutzer and Hannah de Vries, who develop with agents themselves every day. We also show where agents cost time instead of saving it: existing code without tests, unclear requirements, a missing pipeline.
07 — Process
From intro call to certificate. Three steps, one point of contact.
01
Intro call
We talk about your stack, tools, prior experience and what currently causes friction in daily work. Then we select the building blocks and clarify laptops, licences and approvals with your IT.
02
Workshop
One or two days on site at your company. Short theory blocks, followed by exercises on the sample CRM or on your own code. From six participants, two trainers are in the room.
03
Wrap-up
All participants receive a certificate. What was created in the workshop – rules, skills, verification steps – stays in your repository and keeps working the next working day.
Optional
Follow-up support and consulting
On request, we support the first weeks afterwards: we look at how rules and verification steps work in daily practice and improve them. If you want to change your development process as a whole, you will find that in our AI consulting.
08 — For large companies: cooperation with atra.consulting
Agentic Engineering training, also bookable as a fixed format.
We run the Agentic Engineering series together with atra.consulting . If atra is already listed as a service provider at your company, procurement runs through the existing framework – without setting up another supplier. In terms of content, these are the same building blocks, just in a fixed form.
- Advertised by atra for 6 to 12 people, in German or English
- Digital certificate of attendance with content, date and trainers
- Prerequisites: ability to read code, basics in terminal and Git, laptop with Node.js and Git
- For consultants without a developer role, there is also a separate format of 2 × 4 hours
09 — Frequently asked questions
What CTOs and team leads want to know up front. Costs, tools, data protection.
What does AI training for our development team cost?
After the intro call, you receive a fixed-price quote for the entire workshop. The price depends on duration, group size and whether one or two trainers are in the room. If the formats are booked through atra.consulting, the terms published there apply.
Which tools do you cover?
The focus is Claude Code; on request, we also work with GitHub Copilot. The approaches – rules in the repository, specification first, quality gates – transfer to all common providers. For Claude Code, there is also a separate, more in-depth training.
Do we work on our own code, and what about data protection?
By default, we provide a sample repository with prepared exercises. We are also happy to work with your own code. In that case, we clarify in the intro call which repositories and which model providers your approvals allow. We work with your company’s own accounts; production data is not required.
What prior knowledge do participants need?
Programming experience in any language, plus basics in Git and on the command line. Experience with AI tools is not required. If some participants already have a lot of practice, we take that into account when selecting the building blocks.
Is the workshop also possible remotely?
The workshop takes place primarily on site at your company, because that lets us support participants directly on the code and much more closely. Talk to us if your team works in a distributed way.
Our team can teach this to itself. Why a workshop?
You can, and many teams do. But the documentation explains features, not how a team agrees on shared rules, verification steps and responsibilities. That is exactly what we bring from our own projects. You save the months in which each person finds out on their own what works.
You can read more about working with coding agents in our blog.
Request a workshop for your team. Afterwards you will know which building blocks fit.
Tell us briefly how your team works, with which stack and which tools. Or book an intro call directly – afterwards you will know which building blocks fit and what the workshop costs.