What is the most important aspect of live prompting? What are the most important points to look out for?

Dr. Daniel Stoller-Schai: Live prompting means working with AI systems in real time to achieve immediate results. Two aspects are particularly important:

• Precision and transparency: the prompts must be precisely expressed to achieve the desired results. The clearer the instructions, the more relevant the results.
• Flexibility and adaptation: Interaction with the AI requires adaptability - depending on the output, the prompt may need to be adapted directly to optimize the desired results. However, there are prompts that can take care of this.

In the workshop, we will perform practical exercises and use a prompting sheet that helps to create structured and effective prompts more quickly. The focus is on simplifying the input processes and enabling participants to interact confidently with the AI.

Are there specific areas that require special precautions?

Dr. Daniel Stoller-Schai: Yes, according to the new EU AI Act, AI applications are classified into different risk categories:

• Minimal risk: e.g. AI-supported chatbots without security-relevant functions.
• High risk: e.g. AI in medical diagnosis, credit scoring or personnel recruitment.
• Unacceptable risk: Applications that violate fundamental rights, such as certain surveillance practices.

Depending on the area of application, companies will need to take specific measures - from a simple labeling requirement to extensive testing and documentation obligations. However, there will be room for interpretation, which is why it remains important to continuously review the regulatory situation.

What will the EU's new AI regulations change in terms of practical use?

Dr. Daniel Stoller-Schai: The new regulations will focus increasingly on the ethical use of AI and define clear limitations, especially for sensitive applications. Businesses will have to fulfill transparency and verification obligations for certain applications in future.

However, I hope that the scope for innovation will be maintained, as I believe this is important for the European economic area - and Switzerland is part of this.

Europe can also position itself as a pioneer for trustworthy AI - an advantage for companies that want to comply with ethical standards.

However, the exclusion of military applications remains critical, which can be seen as a contradiction to the ethical guidelines. Overall, however, the regulations provide a clear framework that is intended to reconcile innovation and responsibility.

Does it make a difference in prompting which AI I communicate with?

Dr. Daniel Stoller-Schai: Yes, that is a determining factor. Different Large Language Models (LLMs) have different strengths based on their training data, architecture and objectives:

• GPT models (like ChatGPT) are versatile, focusing on language understanding, creativity and text generation.
• Claude (Anthropic) emphasizes security and ethical guidelines.
• LLaMA (Meta) is often used for research and open-source projects, with high customization potential.

Depending on the model, the results, interpretations and even the efficiency of the collaboration can differ significantly. Therefore, it is worth experimenting with different models in a targeted approach to find the right tool for the respective task. There are also platforms such as Straico that allow you to run the same prompt in parallel through different models and then compare the results.

Are there guidelines, instructions or even courses that can provide guidance for users, or does “practice does the trick”?

Dr. Daniel Stoller-Schai: Both play an important role. There are numerous guides, online courses and tutorials - from beginner to expert level. These teach the basics of prompting, provide best practices and help to avoid typical mistakes. The importance of Agentic AI and AI workflows will become apparent in 2025. There are already numerous training courses available.

However, the decisive factor is still practicing. You learn how AI acts, and which adjustments bring the best results through continuous practice and observation.

An additional tip: My preferred form of learning - collaborative learning or peer learning - plays a major role: those who regularly document their experiences and exchange ideas with others (e.g. in forums or networks) will make faster progress and develop their own personal style. LEARNTEC 2025 is a wonderful opportunity you should use.