A few years ago, news broke in the media: Major tech companies were paying salaries that were considered astronomical at the time for a job role that was still largely unknown: \u201cprompt engineer.\u201d AI was just at the beginning of its hype cycle, but one thing was already clear: whoever could use AI most effectively would gain a competitive advantage.<\/p>\n\n
This thesis still holds true. At the same time, the role of the prompt engineer no longer seems as central as once thought \u2013 at least from the user\u2019s perspective. The new large reasoning models and agents, in particular, are getting better and better at discerning the intended meaning. As a result, they usually provide a usable answer even when the prompt isn\u2019t formulated quite perfectly.<\/p>\n\n
The more accessible AI power becomes, the more the question arises: What are some concrete examples of how AI can actually be used right now? Perhaps it\u2019s comparable to the situation when you\u2019ve just passed your driver\u2019s license test and then think, \u201cCool, but where do I go now?\u201d<\/p>\n\n
Here\u2019s an example from everyday work life: Scheduling a 45-minute meeting with the topic \u201ccoordination\u201d and no further details won\u2019t lead to any increase in effectiveness, even with the use of AI. Not because the task is too complex for AI, but because no task has been defined at all \u2013 only the intention to come up with one later. The prompt \u201cCreate an agenda for a 45-minute coordination meeting with the goal of agreeing on the most important deliverables for the first project phase. Find a time slot in Sarah\u2019s calendar and invite her to the meeting\u201d<\/em> describes the desired outcome and exactly what is expected of the AI. <\/p>\n\n This is precisely where a parallel emerges that, at first glance, seems unrelated but essentially describes the same mechanism: the parallel between coaching and prompt engineering. A good coach doesn\u2019t solve their coachees\u2019 problems themselves. Instead, they help transform a vague sense of unease (\u201cI\u2019d like to be fitter\u201d) into a clear, actionable goal (\u201cIn 6 months, I want to be fit enough to finish a half-marathon feeling good\u201d). Only once this goal is established, a path to achieving it can be planned and worked toward.<\/p>\n\n Prompt engineering works on the same principle. An AI can structure text, write code, weigh options, prepare decisions and even execute tasks \u2013 but only based on a goal that is described precisely enough to derive concrete steps from it. The AI is thus less the coach itself and more of an extremely powerful execution layer that only unleashes its full potential when given a clear description of the goal. The better the goal is described, the more direct and valuable the result. Conversely, an unclear goal description also increases the likelihood of heading in the wrong direction.<\/p>\n\n The central question is therefore not (just) \u201cHow do I write the perfect prompt?\u201d but \u201cWhat exactly do I actually want to achieve?\u201d \u2013 a question that is gaining importance as AI capabilities continue to advance.<\/p>\n\n Of course, there\u2019s also a buzzword for this topic: \u201cIntent Engineering.\u201d<\/p>\n\n A clearly formulated goal is a necessary prerequisite, but it is not sufficient on its own. This is because there is a second, often underestimated step between the goal and the result: the ability to ask the right questions in a targeted manner in order to get the right answers. If you know your goal but ask the wrong questions, you still won\u2019t get what you actually need.<\/p>\n\n Questioning techniques are a key focus in many coaching training programs \u2013 not answering questions, but asking questions that truly prompt the other person to think. A coach aims to activate problem-solving in the coachee. A typical question might be, for example: \u201cWhat would need to happen for you to feel satisfied with the result in three months?\u201d This phrasing elicits a more substantive answer with suggestions for achieving the goal, rather than just an excuse for why it wasn\u2019t achieved.<\/p>\n\n In prompt engineering, the effect is identical: Two prompts with the same goal can yield completely different results, depending on how precisely they\u2019re phrased, what follow-up questions they allow, and what parameters they set. A prompt like \u201cCreate a sales plan for me\u201d yields an initial proposal that may even look promising, but it doesn\u2019t necessarily have to fit your own company. A prompt like \u201cWhat three intermediate goals do we need to achieve in order to reach our new sales target of 2 million euros yearly turnover in our department? Create a plan for achieving these goals, including measurable metrics so we can continuously track our progress\u201d prompts the AI to engage in the same kind of substantive thinking that a good coach elicits from a client.<\/p>\n\n The three components thus form a cohesive triad: Coaching provides the mindset and process for refining a vague intention into a clear goal. Questioning techniques provide the tools to test this goal, reveal assumptions, and get to the heart of the matter. Prompt engineering applies these principles to interactions with AI. Those who master all three levels not only get answers faster but also increase the likelihood that those answers align with the actual goal.<\/p>\n\n Anyone involved in projects is familiar with the moment when a client says, \u201cWe need a better solution for X.\u201d However, the actual consulting work only begins when this statement is transformed into a concrete, measurable goal description. It is precisely this translation step where consultants can draw on their experience from the various companies and projects they\u2019ve already worked on. So, by combining methodologies for defining specific goals with effective questioning and the consultants\u2019 ability to fill in the unspoken assumptions, AI can ultimately be used truly effectively.<\/p>\n\n This also creates more parallels between consultants and coaches: a precise understanding of the client\u2019s requirements has always been \u2013 and will become even more \u2013 central. Through truly good, open-ended, and results-oriented questions, the exact goal is jointly formulated in concrete terms (ideally using the SMART criteria: Specific, Measurable, Achievable, Realistic, Time-Bound).<\/p>\n\n This lays the foundation for the subsequent \u201ctechnical\u201d phase: using technology in the best possible way to achieve this goal \u2013 and, more recently, increasingly with the help of AI. <\/p>\n\n Efficiency gains through AI do not result directly from the use of newer and more powerful programs and models. AI is a powerful tool \u2013 there\u2019s no question about that. However, to harness its true potential, a concrete goal is needed, followed by precise instructions on what the AI should do: The use of AI is only as effective as the goal behind it \u2013 technology alone cannot resolve vague intentions.<\/p>\n\n In summary, therefore, we can state:<\/p>\n\nA look at Prompt Engineering<\/h3>\n\n
From the goal to the right question \u2013 The Second Skill<\/h2>\n\n
The right question in Prompt Engineering<\/h3>\n\n
Specific Goal + Effective Questions = Added Value Through AI<\/h2>\n\n
Conclusion<\/h2>\n\n