Reference architecture for AI prompts

The rapid rise of Generative Artificial Intelligence is transforming the way we interact with technology. The effectiveness of these interactions depends on the quality of the “prompts” – the instructions we give to AI models. This whitepaper describes a reference architecture for AI prompts. This goes beyond simply writing good prompts; it provides a structured, systematic approach to designing, managing and optimizing interactions with AI systems within an organization.

Why is an architectural approach to prompting relevant? In an enterprise context, consistency, reusability, governance and scalability are crucial. Without a well-thought-out approach, the results of AI interactions and the performance of AI models will remain variable and difficult to optimize at the organizational level. After all, the effectiveness of AI does not only depend on the AI ​​model itself, but increasingly on the human-machine interface, in which the quality of the prompt is central. Architecture provides the principles and structures to standardize and improve this interaction. The absence of such a prompt architecture can lead to “AI sprawl” within organizations, with inconsistent results, incremental investments, duplication of effort, and potential risks such as bias and misinformation. This undermines the strategic value of AI investments. An architecture positively influences the implementation of AI and aligns it with the organization’s capabilities and the intended level of innovative progress.

Read whitepaper (in Dutch)