Translated by an automated translation plugin.
Good prompts improve results: Clear instructions specifying the role, context, task and desired format help to use generative AI in a more targeted and efficient manner.
AI literacy does not require programming skills: Staff from departments such as marketing, human resources or finance can make effective use of AI applications with their specialist knowledge.
Shared prompt libraries add value: Organisations can centrally collect, review and reuse tried-and-tested instructions. This promotes knowledge sharing and more consistent results.
Fewer iterations conserve resources: Precise prompts can reduce token consumption, thereby lowering the energy and resource requirements of AI applications.
Training ensures long-term benefits: Role-specific training helps to embed prompt engineering as a shared skill within the organisation.
In the seventh instalment of the series ‘Switzerland as a Hub for Innovation’, Dr Lamia Friha, Prof. Dr Giovanna Di Marzo Serugendo, Dr Pierre-Yves Burgi and Jan Melichar from the University of Geneva explain how companies can make targeted use of generative AI through prompt engineering. The article explains what makes a good prompt, how various techniques can be applied in day-to-day work, and why the sharing of tried-and-tested prompts offers a strategic advantage, particularly for SMEs.
The authors also explain how shared prompt libraries can improve the quality and consistency of AI outputs. Another key focus is on the responsible use of generative AI: using the open-source learning tool ‘BrAIn Ecosystem’ as an example, they demonstrate how resource consumption can be made visible and how targeted prompt design can contribute to a more sustainable use of AI.
This article in KMU-Magazin forms part of the series ‘Switzerland as a Hub for Innovation’, which emerged as a follow-up to the SATW publication ‘AI Orientation: Challenges and Opportunities for Swiss SMEs’.
The content was developed in collaboration with stakeholders from the SAIROP (Swiss AI Research Overview Platform) network and the SATW’s network of experts. SAIROP promotes collaboration between research, industry and public administration and supports knowledge transfer within the Swiss AI ecosystem.
Prompt engineering refers to the structuring of instructions for generative AI systems. By providing clear task descriptions, background information and guidelines on the desired response, it is possible to achieve more relevant and useful results. For SMEs, this offers the opportunity to deploy AI across various business areas even without specialist development teams.
A good prompt describes the role of the AI, the relevant context and the specific task. In addition, the desired output format, tone and specific constraints can be specified. Examples help to clarify the expected results. The more clearly the requirements are formulated, the more precisely the AI system can respond. The results must, however, be reviewed by an expert.
In a shared prompt library, organisations can centrally store, review and make best-practice prompts available to other staff members. This makes it easier to reuse successful prompts, encourages the sharing of knowledge and improves the consistency of results. Regular quality, security and compliance checks help to ensure the library is developed reliably.
Precise and well-structured prompts can reduce the number of follow-up queries and corrections required. This lowers token consumption, which can also reduce resource requirements. The open-source tool ‘BrAIn Ecosystem’, developed by the University of Geneva and its partners, highlights the estimated energy consumption, water usage and carbon footprint of AI applications and helps users improve their prompts.
A sensible starting point is practical training tailored to different professional groups. Staff can start by identifying suitable use cases, testing prompts and critically reviewing the results. Proven instructions can then be collated in a shared library. It is crucial to update this knowledge regularly and to encourage knowledge-sharing between teams.
| Role | Title + Name |
|---|---|
| Text by | Pierre-Yves Burgi, Giovanna Di Marzo Serugendo, Lamia Friha |
