Why I Started Thinking About AI Skills

Like many SAP developers, I started using AI to make my daily work easier. Sometimes it helped me write a CDS model. Sometimes it explained an API that I wasn't familiar with. Sometimes it saved me from spending hours debugging an issue.

The more I used it, the more useful it became.

But after a while, I noticed something.

Most of my work wasn't about writing code.

It was about following a process.

When I deploy an application to SAP BTP, I don't just run a deployment command. I first check the target. I verify that I'm deploying to the correct space. I make sure the required plugins are installed. I confirm the deployment strategy. After deployment, I verify the application, check the services, and review the logs.

None of that is a single prompt. It's a workflow that experienced developers follow almost without thinking. That's when I started wondering whether AI should learn workflows instead of simply answering questions.

Prompts Are Good. Workflows Are Better.

A prompt is great when you want an answer to a question.

“Explain this error.”

“Generate a CAP entity.”

“Write a SQL query.”

Those are all useful.

But enterprise development is different.

Most engineering tasks involve multiple steps. Some steps need validation. Some require confirmation. Some can be dangerous if executed incorrectly.

For example, deploying an application to the wrong Cloud Foundry space can affect other developers or even a production system.

In those situations, I don't want AI to simply give me the next command.

I want it to follow the same process that an experienced SAP developer would follow.

That's When the Idea of an AI Skill Came In

Instead of asking AI to answer a question, what if we asked it to perform a complete engineering task?

Not by taking control, but by guiding the developer through the entire workflow.

That's how I started thinking about AI Skills.

An AI Skill has a single purpose.

It knows what it is responsible for.

It follows a defined workflow.

It validates before acting.

It explains what it's doing.

It helps the developer make the right decision instead of simply generating commands.

In other words, an AI Skill behaves more like an experienced teammate than a search engine.

The Principles I Want Every Skill to Follow

This is the most important section in the article.

I would make it a simple list.

1. One Skill. One Responsibility.

Every skill should solve one problem well.

A deployment skill should focus on deployment.

A project validator should validate projects.

A troubleshooting skill should troubleshoot.

Keeping skills focused makes them easier to use and easier to maintain.

2. Follow a Workflow

Enterprise development is a sequence of steps.

AI should guide developers through those steps instead of jumping directly to the final answer.

3. Safety First

Some operations are harmless.

Others can change systems.

Some are irreversible.

Every skill should understand the difference and behave accordingly.

4. Explain Every Decision

Developers should always understand why a recommendation is being made.

AI should never feel like a black box.

5. Stay Within Scope

A deployment skill should not suddenly become a security advisor.

A HANA skill should not review UI code.

Every skill should have clear boundaries.

This Is the Journey I'd Like to Share

This article is the beginning of a series called AI Skills for SAP BTP.

Over the next few weeks, I'll be sharing the skills I'm building, the design decisions behind them, and the lessons I learn along the way.

The goal isn't to replace developers.

The goal is to capture good engineering practices and make them reusable.Some of the skills I'll be covering in this series include:

 

🚀 Deployment Skills

• Cloud Foundry Deployment Assistant – Guides developers through a safe and structured deployment process with validations, confirmations, deployment verification, and troubleshooting.

✅ Project Quality Skills

• CAP Project Validator – Reviews SAP CAP projects and identifies configuration issues, missing dependencies, and common development mistakes before deployment.
• SAP BTP Compliance Analyzer – Checks applications against SAP BTP recommendations, security practices, deployment readiness, and platform configuration.
• SAP BTP Best Practices Analyzer – Reviews projects against SAP development standards, coding guidelines, naming conventions, and recommended design patterns.

🔍 Troubleshooting Skills

• Deployment Troubleshooter – Helps developers understand deployment failures, identify the root cause, and recommend possible fixes.

As the series grows, I'll also explore skills for other areas such as SAP HANA Cloud, SAP Integration Suite, SAP Fiori, and other common SAP BTP development tasks.

I hope these articles are useful whether you're just starting with SAP BTP or you're an experienced developer looking for ideas on how AI can support day-to-day work.