The annual “AI for SMEs” conference organised by the Chamber of Commerce and Industry for Munich and Upper Bavaria offered practical insights into how companies are using AI. One particularly interesting session was a ChatGPT workshop run by the AppliedAI Institute, which focused on prompt engineering and context engineering as key methods for improving research.
Prompt engineering helps guide the AI through clear instructions. This includes precise requirements for wording, style and structure, as well as a clearly defined task and role. Context engineering adds another layer by ensuring that the AI gathers relevant background knowledge before the actual prompt is given. When planning an AI event, for example, you might first ask strategic questions about recent developments in AI or influential figures in the field before asking the model to draft a plan for the event.
The workshop also emphasised the importance of keeping chat histories focused. New topics should ideally be started in new chats, as the model primarily works with the context of the current conversation. In very long chats, ChatGPT tends to give greater weight to the most recent prompts and answers – known as the recency effect. At the same time, the very first prompt can also have a lasting influence, particularly when it contains a clear role or task definition – known as the primacy effect.
By Maria Kleiner
