Skip to content

Mastertutorial · Part 4

Where does artificial intelligence belong in IT systems?

With Christian Liebich19 minArtificial intelligence

In short

Artificial intelligence is a service within the solution architecture and not a replacement for digital transformation. It belongs where data is analysed, patterns are recognised or content is generated, integrated through APIs for a narrowly defined use case. Everything that has to be structured, reliable and verifiable remains the job of classic software, and critical decisions need a human in the loop.

The video is spoken in German and has integrated English subtitles. The podcast episode and the blog article are in German. This page summarises the episode in English.

The video is loaded from YouTube only after you click. Data is then transferred to YouTube.

What you take away

  • AI amplifies digital transformation, but does not replace it.
  • The narrower and clearer the use case, the more reliable the result.
  • Chained AI systems (agents) multiply errors, cost and latency.
  • For companies, small specialised models are often the more strategic choice than large generalists.

Topics of this episode

  1. Hype and reality
  2. Classic software architecture for structured data
  3. Machine learning for data analysis
  4. Generative AI and LLMs: strengths and limits
  5. The EU AI Act and the human in the loop
  6. A critical look at AI agents and agentic AI
  7. Integration through APIs: generalist or specialist
  8. Problem → use cases → solution design

Who it is for

For decision-makers who decide on the use of AI, automation or an AI strategy and want to separate benefit from hype.

This episode as

Related service

AI Strategy & Integration

Where does AI fit your business model, systems and processes, and what does it cost and return? I give you answers without the hype.

Frequently asked questions

Does AI replace digital transformation?

No. Putting AI on top of old processes does not make them more efficient, it makes them unstable. Rethinking processes, breaking up structures and getting data in order remains the real work.

What is classic software better at than AI?

Everything with clear rules and deterministic flows: master data, accounting, logistics, order management, government processes and regulatory documentation.

How is AI integrated into a system in a useful way?

Not by giving every employee a ChatGPT account, but programmatically through APIs for a specific use case. Either with a large model as a generalist or with your own specialised model.

All Mastertutorials

Let us talk about your project

A first conversation takes 30 minutes and is free of charge. Afterwards you know which next steps make sense.