Prompting & AI Tools in Daily Work

Prompting & AI Tools in Daily Work The two-day seminar for everyone who uses AI tools and does not want to guess: you learn how a language model works, why it answers convincingly wrong and how to phrase tasks so that usable results come back. By the end you have five prompts of your own that you need every day.

Prompting & AI Tools at a glance

• Format: 2 days on site + self-learning track
• Level: basic
• Participants: max. 16 people
• Language: German (English on request)
• Certificate: AEI certificate of attendance
• Investment: €99.00 plus VAT

Most people use AI like a search engine: one question, one answer, and if it does not fit, the tool is to blame. In fact the task — what you put in — almost alone decides the quality of what comes out.

The second misconception is more dangerous. A language model always answers fluently, even when it knows nothing. Whoever does not distinguish that copies invented figures and false sources into documents that leave the company.

This seminar works on both points. You understand what a model does and where its limits are, and you practise phrasing tasks so that result and expectation match. You practise on your own texts, not on examples.

What you take away

• An understanding of what a language model calculates — and what it does not.
• A prompt framework of four parts that you can apply to any task.
• A technique for better results: examples instead of explanations.
• A routine for checking results before they are reused.
• Five prompts of your own for the tasks you have every week.

Your learning objectives

• You explain why a language model answers fluently without knowing.
• You recognise hallucinations before they reach a document.
• You phrase role, goal, context and format in one prompt.
• You improve results through examples and iteration instead of repetition.
• You decide which data may go into a prompt and which may not.

The levers at a glance

The model calculates probabilities, it does not look things up. It says what most probably follows the text so far. That explains both the astonishingly good phrasing and the invented citations.

One example does more than three explanations. Whoever shows what the result should look like gets it more often. Whoever describes what they want gets a description.

Iteration beats the perfect first prompt. The first draft is a starting point. Whoever sharpens three times gets further than someone who starts over five times.

Data protection is decided in the prompt. What you type leaves your device. Customer names, contracts and personal data do not belong in a third-party service — regardless of what the tool promises.

A model has no memory of your company. It does not know your products, your tone or your prices. What it knows is the pattern of language. So the context you supply is not politeness — it is the whole difference between a generic sentence and a usable one.

Who is this seminar for?

For people who use AI tools in daily work or want to introduce them.

• Marketing and content teams who have to produce texts faster.
• E-commerce and shop managers who write product texts and replies.
• Assistants and service staff who handle many similar requests.
• Leaders who need to assess what AI can do today.

No prior knowledge is required. We work with common tools; the way of thinking is tool-independent and remains valid when providers change.

The seminar is not suitable for developers who want to train their own models or connect an API. We work on application, not on building.

Investment and dates

Participation costs €99.00 plus VAT, including the self-learning track, materials and certificate of attendance. There are six dates per year in Berlin, Hamburg, Cologne, Munich and live online, plus an in-house variant for teams. You will find all dates in the list on this page.

For the exercises you need a device with internet access and a free account with one of the common providers. We tell you which one before the date.

After the seminar

You keep the materials, the prompt framework as a one-pager and your five prompts in a file you open daily. That is the real yield: not knowledge about AI but a collection that saves work.

Keep the file short. Five prompts that you actually use beat forty that you collected and never opened, and a short list is one you can still read in six months.

Two weeks after the seminar you receive a short follow-up: one task you solved with your framework and the option to look at the result with us.

For teams the seminar is available as an in-house event with your own texts and tasks. The procedure stays the same, the examples come from you.

Programme: 2 days

Day 1 – Fundamentals: how AI language models work

• What a language model really does
• Recognising limits & hallucinations
• Data protection in the prompt
• Role, goal, context, format
• Examples instead of explanations (few-shot)
• Iterating instead of asking once

Day 2 – AI tools in daily work

• Working with text
• Images & presentations
• Research with sources
• Product texts
• Customer service replies
• Evaluating feedback

Both days run from 09:00 to 14:35 with breaks in the morning and at midday.

Each day has three parts: a short introduction to the pattern, an exercise on your own task and a review in the group. At the end of day one you phrase one task that you want to solve completely on day two.

The seminar includes a self-learning track. It contains the patterns from both days as short units with examples and one self-check each. You can work through it before the date to clarify terms, or afterwards to repeat what stuck and what did not.

Our approach

We work on your tasks, not on a collection of tricks. At the start of each exercise you decide which task you want to solve and how you would recognise that the result is usable; at the end stands a prompt you can reuse.

The exercises are short and repeat in variations. That is deliberate: prompting is not learned by watching but by sharpening repeatedly on real material.

Every block ends with a sentence you could show a colleague: which task, which prompt, and how you would notice that the model invented something.

We work in pairs, in small groups and in plenary, with changing partners. Every participant gets feedback on their prompt, not just on their idea. Group level and participant count are deliberately limited: with 16 people no anonymous lecture is possible.