AI automates mental work - from simple chatbots to agentic AI. The article shows how you can build up AI skills with learning by doing and why automation is becoming a key skill in continuing education
AI: hype, craftsmanship and homework - why everyone is now talking about automation
It’s hard to avoid it in the media and in everyday working life at the moment: artificial intelligence. In the office, a colleague says that he is having his presentation "quickly built by ChatGPT". People post pictures on LinkedIn that have never been photographed. And Microsoft, Google & Co. are installing "co-pilots" everywhere that can supposedly do everything better and faster - from emails to Excel formulas. The good news: behind all the hype is something pretty down-to-earth. At its core, AI is nothing more than automation - just for mental work instead of muscle work.
AI transformation: Formerly an assembly line, now a desk - and a maturity model
Automation is nothing new. Machines used to do physical work: Assembly lines, robots, manufacturing. Today, it’s mainly desk-based tasks: writing, sorting, planning, analyzing, checking. The maturity of an AI application is shown by the tasks it takes on. This can be read as a maturity model - from a simple gimmick to an orchestrated AI organization (see illustration).
At the very beginning is the simple LLM: a classic question-and-answer system. You type something in and the model responds - without any knowledge of the company, just on the basis of its pre-trained model.
The next step is the RAG approach (Retrieval Augmented Generation): A separate knowledge base is added to the pure AI, a small collection of documents for example. The AI first retrieves relevant documents from a database and formulates an answer based on them. "Just chat" becomes: Question - search - answer. This is followed by agents: AI components that not only answer, but also actively take on tasks - such as retrieving data, creating files or initiating workflows. In a multi-agent system, several such agents work together, share tasks and pass on intermediate results. The highest level is Agentic AI: a kind of AI ecosystem in which several agents work together in a coordinated manner, processes are optimized over time and new solutions are "discovered". Here, we are no longer talking about individual tools, but about learning, orchestrated AI structures that help shape entire workflows
Learning by doing: AI is not something you learn ’from slides’ or by reading ten articles.
This is where it gets exciting for training participants: AI is not a topic that you can just understand - you have to apply it and ’grasp’ it, so to speak. You don’t become AI-competent by reading ten articles. You become AI-competent by trying out ten specific tasks with AI, for example.1. You enter a task ("Answer the following email to Ms. Stohler...").
2. The result is "okay, but not yet good".
3. You become more precise: add context, target group, tonality, length.
4. You realize: the clearer the brief, the better the AI.
This loop - try, adapt, repeat - is the actual learning by doing. With each attempt, you not only learn how the AI reacts, but also how well (or badly) you can describe your own tasks. In this sense, AI is not just an automation tool, but also a mirror: if you don’t understand your own processes, you won’t be able to automate them well.
If you want to have a say in the future when it comes to AI, you don’t necessarily need a technical background, but rather a technical affinity. But they do need a good understanding of processes and tasks. This turns AI expertise into automation expertise: the focus is not on the question of which tool is hip at the moment, but on the question of how I can set up my work in this way: How can I set up my work in such a way that AI helps me in a meaningful way?
Accordingly, the continuing education formats at Lucerne University of Applied Sciences and Arts are also practice-oriented - such as the CAS Business and Service Innovation. Here, participants not only learn about service innovation and new business models, but also learn how to use AI in their own projects.
