ABAP development
with AI agents
In two intensive days, your development team learns to use Claude Code, GitHub Copilot and MCP servers productively – not in a demo environment, but on your SAP system and your own code.
Who is this training for?
For SAP teams who don't just want to experiment with AI, but want to make it a reliable part of their everyday development work.
ABAP developers
You know ABAP, CDS and your custom Z landscape – and want to learn how AI agents help you analyse, refactor and build new applications faster.
Core audienceLead developers & architects
You set the standards. Here you develop the rules, prompt templates and review processes that let the whole team work with AI consistently and safely.
Standards & qualityDevelopment & IT management
You need a solid assessment: where does AI really save time, which tools fit your landscape, and what do you need to consider regarding data protection and governance?
Basis for decisionsWhy AI isn't taking off in many SAP teams
The AI doesn't know your system
Without access to your Z objects, data model and naming conventions, even the best model hallucinates – and generates code that never activates.
Everyone experiments alone
Some use ChatGPT, others Copilot, many nothing at all. There are no shared tools, no rules and no common understanding of what actually works.
Uncertainty slows things down
Can our code be sent to a language model? Who reviews the result? Open questions on data protection and quality keep AI out of real projects.
A licence alone doesn't make an AI-ready team.
“The demo looked great. Unfortunately it had nothing to do with our 15-year-old custom code.”
Hand out licences and hope
Without onboarding, the AI assistant remains a slightly better autocomplete. The potential of agentic ways of working is never realised.
Tutorials & videos
Most examples come from web development. SAP specifics such as transports, ADT, RAP or authorisations barely feature.
Standard courses with someone else's examples
Exercises on a training system are a good start. But the hard part – transferring it to your landscape – is left to each individual.
What it really takes – and what we bring
What your team needs
- A secure setup that connects AI to the SAP system
- A way of working: specify, delegate, review
- Exercises on your own code instead of toy examples
- Shared rules, prompts and context files
- Clear guardrails for data protection and code review
- Someone who knows SAP and AI from real projects
Consetto's in-house training delivers exactly that
- A working setup by the end of day 1Claude Code or Copilot plus MCP server – installed on your machines and connected to your development system.
- A new way of working, practised hands-onFrom single prompts to precise specifications that an agent implements reliably.
- Your use cases as exercisesWe agree in advance which packages, reports or dumps we will work on together.
- Team standards to take awayContext files, prompt library and review checklist – created for your team.
- Governance from the startModel choice, data flows, authorisations and quality assurance are discussed openly.
Two days. From the first prompt to a team standard.
Day 1 lays the foundation and delivers a working setup. Day 2 applies AI to real tasks from your everyday development. We adjust the emphasis to your priorities in the preparatory call.
Live kick-off: the developer's new role
We start without slides: together we build a first application in a few minutes – and discuss what this changes about the way we work.
- From code completion to autonomous agents
- Specify, delegate, review – the new work cycle
- Realistic expectations: strengths and limits of today's models
Context engineering instead of prompt luck
Good results depend less on the individual prompt than on the context the agent receives.
CLAUDE.mdAGENTS.mdSpecifications- Project rules, naming conventions and architecture guidelines as context files
- Writing specifications that an agent implements reliably
- Prompt patterns compared side by side – what really makes the difference
The tool landscape for ABAP
Which tools are available, how do they differ and which one suits your team?
Claude CodeGitHub CopilotEclipse ADTVS Code- Claude Code in the terminal, Copilot in Eclipse ADT and VS Code
- Where Joule for Developers fits: when SAP's own AI, when external agents?
- Working locally with abapGit export vs. direct system access
MCP: the AI talks to your SAP system
With the Model Context Protocol, the agent reads and understands objects directly in the system. By the end of the day, the setup runs for every participant.
MCP serverADT interface- Setting up an ABAP MCP server, e.g. ARC-1, against your development system
- Starting safely: read-only mode, package filters and authorisations
- Hands-on: analyse a package and generate documentation
Understanding and modernising legacy code
The biggest lever in grown landscapes: extracting knowledge from old code and making it future-proof step by step.
RefactoringClean Core- Analysing complex custom applications – dependencies, data flows, risks
- Refactoring old reports and fixing ABAP Cloud findings
- Knowledge transfer: generating documentation for onboarding and maintenance
Building RAP applications with AI
From describing a business object to a Fiori Elements UI – including tests.
CDSBDEFABAP Unit- Generating table, CDS views, behavior definition and service binding
- Implementing determinations, validations and actions
- Creating ABAP Unit and CDS tests and letting the AI work against them
Finding errors faster
When things break, speed matters. The agent gathers the evidence, you make the call.
DumpsLogsTransports- Analysing runtime errors and logs to narrow down root causes
- Investigating performance issues in SQL and CDS
- AI-assisted review of transports and changes before release
Governance, outlook & team roadmap
We close with what matters after the training: binding rules and a concrete plan for your team.
Data protectionMulti-agentRoadmap- Model choice, data flows and data protection – what goes where?
- Outlook: custom MCP servers, AI calls from ABAP and multi-agent workflows
- Created together: team rules, review checklist and next steps
How your in-house training works
The two days on site are the core – the real impact comes from good preparation and a smooth transition into everyday work.
Preparatory call
We clarify goals, prior knowledge, system landscape, preferred tools and your data protection requirements.
Setup & use cases
A checklist for laptops and system access. Together we select packages and tasks from your landscape for the exercises.
Two training days
Short inputs, lots of hands-on work. Everyone works on their own laptop – with your system and your code.
Transfer into daily work
You receive materials, context files and a prompt library. On request, we support the first weeks with follow-up sessions.
Your system. Your code. Your team.
Exercises on real objects
Instead of a training system, we work with your development or sandbox environment. What you learn on Tuesday, you use in your project on Wednesday.
The whole team on the same page
Learning together means shared terminology, tools and rules. Ideal for 3 to 15 people – small enough for individual support, large enough for real exchange.
Confidential
Your code, your architecture and your weak spots stay within your own four walls.
Tailored
We weight topics such as RAP, Clean Core or error analysis according to your needs.
Easy to plan
Dates that fit your calendar, no travel costs for the team, no waiting for public course dates.
One package. Everything your team needs.
ABAP Development with AI
2 days on site · for teams of 3 to 15 people
Tailored to team size, location and focus topics.
- Preparatory call and agreement on exercise cases
- Setup checklist and support with preparation
- Two training days on site at your company
- Hands-on work on your own system with your objects
- Training materials, context files and prompt library
- Team rules and review checklist as outcomes
- Optional: follow-up sessions to support your projects
Prerequisites
- Solid ABAP skillsNo prior AI knowledge required.
- Laptop with Eclipse ADTOptionally VS Code and Node.js – you'll receive the checklist in advance.
- Access to a development or sandbox systemIdeally with objects your team actually works on.
- Access to AI toolsYour own licences or trial access: we clarify this in the preparatory call.
- A training room with projector and stable networkAt your premises or a location of your choice.
Built on projects, not slides.
Since 2014, Consetto has supported companies with SAP development, analytics and Clean Core transformations. We use AI agents every day in our own customer projects – and we build the tools for it ourselves.
Frequently asked questions
Can't find your question? Get in touch – we usually reply within one business day.
Ask a questionDo participants need prior AI experience?
No. What matters is solid ABAP knowledge. We build up everything around AI tools, prompting and MCP step by step during the training. Participants with prior experience benefit especially from day 2 and the team standards.
Will our source code be sent to an AI model?
When working with cloud models, parts of the code are sent to the provider as context. Which providers, contract models and settings are suitable for you is something we clarify in the preparatory call. In the training we show how to make data flows transparent and restrict access, e.g. via read-only mode and package filters.
Which SAP systems are supported?
You need a system you can work with via Eclipse ADT – this applies to SAP S/4HANA as well as many ECC systems. The exact possibilities depend on release and configuration; we check this together beforehand.
Which AI tools do we use?
The focus is on Claude Code and GitHub Copilot combined with MCP servers for SAP. If you already use other tools or would like to use SAP Joule for Developers, we take that into account in the planning.
We have more than 15 people – what now?
To make sure everyone can participate actively, we limit the group size to 15. For larger teams we offer several sessions – also with different focus areas, for example for development and architecture.
Can the content be customised?
Yes, that's the core of the in-house format. The eight modules provide the framework; we align emphasis, examples and exercises with your goals – for example more RAP, more legacy analysis or a stronger focus on governance.
Is the training also available remotely?
The training is deliberately designed as an on-site format: setup issues get solved faster in person, and discussions about shared team standards work best in the room. Shorter follow-up sessions after the training can of course take place remotely.
Bring AI into your ABAP team.
Tell us briefly about your team and your goals. We'll get back to you to arrange a no-obligation preparatory call.