The first violin
Imagine you book an orchestra. 80 musicians, every instrument played, from double bass to piccolo. Then comes the big evening... and you only let the first violin play.
This is exactly what is happening in most companies with AI.
You license GPT-4, Claude or Gemini and pay for Enterprise plans, for API access, for custom instructions. And then you are asked (PROMPTed) something like this: „Summarize this text.” „Write me an email.” „Make this shorter.”
It works. Nobody disputes that. But it's a fraction of what these models can actually do. Not because the models secretly have more to offer. But because The way in which we address them determines which capacities are activated.
This is not prompting wisdom but architecture.
How a language model actually works
To understand why the quality of your question so fundamentally affects the quality of the answer, it helps to look at what happens in a language model when it processes a prompt. Not at PhD level, but deep enough to understand why „Summarize” and „Analyze the strategic implications considering X, Y and Z” not only generate different answers, but activate different processing paths in the model.
Five mechanisms are central to this.
1. attention - who listens to whom
Transformer-based language models - and these are practically all major models today - work with a mechanism called Attention. Simplified: Each word in your prompt „looks” at every other word and decides how relevant it is for the next calculation.
With a short, low-context prompt such as „Summarize”, there is little to focus attention on. The model is given a narrow window and works correspondingly narrowly. If, on the other hand, you give a prompt with context, perspective, target group and specific requirements, a dense network of cross-references is created. The model can create connections that simply do not exist with a three-word prompt.
Imagine Attention as the conductor. With a simple prompt, he only points to the first violin. With a rich prompt, he conducts the full orchestra.
Try it out: Give a model the assignment „Write me an email to a customer” - and the attention spreads thinly over a few generic patterns. Instead, give it „Write an email to the CTO of a medium-sized mechanical engineering company that is skeptical about cloud solutions, using the tone we used at the last trade fair” - and suddenly dozens of context points are linked. Mechanical engineering jargon. B2B tonality. Addressing skepticism. The model works harder because it has more material to work with.
2nd layer depth - how far the thinking goes
Large language models consist of dozens to over a hundred processing layers. Each layer transforms the information a little further - from raw text recognition to more abstract concepts.
Simple tasks are typically „solved” in the early layers. The model recognizes the pattern, generates an answer and the later layers do little to change it. Complex tasks, on the other hand, require the deeper layers: This is where abstraction takes place, context interweaving, weighing up different interpretations.
If you set simple prompts, you are - figuratively speaking - only using half of the instrument. The back layers run along, but do not make a significant difference. If you ask complex, well-structured prompts, you activate capacities that simply lie idle with trivial queries.
3. prediction entropy - how broadly the model thinks
Language models generate their answers token by token - basically word by word. At each step, the model calculates a probability distribution: Which word comes next?
For simple, predictable tasks, this distribution is peaked. The model is pretty sure what comes next. „The Eiffel Tower is located in...” → „Paris”. Low entropy, low surprise, low computational effort.
For complex tasks, the distribution becomes flatter. Many words could come next. The model navigates through a wider range of possibilities, weighs up alternatives and generates more differentiated formulations.
Simple questions generate simple distributions. The model stays on a narrow track. Complex questions open up the space - and that's exactly where the answers emerge that really help you move forward.
In practice, this means that if your prompt only allows one obvious answer, you will get exactly that. The model chooses the most likely path and is quickly finished. But if your prompt opens up a space in which several good answers are possible - then the model navigates this space. It weighs up, it differentiates, it finds formulations that are not obvious. This is no coincidence. This is statistics reacting to complexity.
4. reasoning chains - when the model thinks out loud
Newer models such as Claude with „Extended Thinking” or GPT-4 and 5 with Chain-of-Thought can generate visible intermediate steps. The model reveals its considerations before coming to an answer.
What happens is technically remarkable: the internal reasoning chain scales with the complexity of the task. Ask a simple question - short thought process. Ask a complex, multi-dimensional question and the model generates longer, more convoluted reasoning chains. Not because it is programmed to write more. But because the task structure requires it.
This is where the orchestra metaphor becomes particularly tangible: The complexity of your score determines how many instruments play. Not the other way around.
5. knowledge networking - narrow corridor or open field
Language models do not store knowledge in discrete entries like a database. Knowledge is distributed across billions of parameters, as patterns, as weightings, as statistical relationships between concepts.
A narrow, specific question activates a narrow corridor of these parameters. The answer comes from a limited area of the model. A broad, context-rich question, on the other hand, activates parameters across different areas of knowledge. The model can make connections that cannot be made with a narrow question because the relevant parameters are simply not addressed.
Imagine it like this: If you ask the price of an instrument, the cashier answers. If you ask about the role of this instrument in the music history of the 19th century, a whole ensemble of knowledge answers - music theory, history, acoustics, cultural studies.
What I have observed
I work with 14 AI agents every day. Different models, tasks and contexts. A curious observation led me to write this article:
The same model - Claude Opus - revised the personality of one of my AI agents in a work session. At the same time, it explained its own behavior. Not on demand. But because the context brought it out. It described why it works differently in this specific task than in a simple summary. Which internal patterns it uses. Why the response structure changes.
Same model. Same license. Same API. Different music.
That was no coincidence. It was the direct consequence of the five mechanisms I described above. The context was rich. The task was complex. The attention patterns were dense. And the model worked accordingly deeply.
What this means for your company
The practical consequence is inconvenient, but important:
If you want better results from AI, you don't need a more expensive model, but a better understanding of what is already there.
This does not mean „10 prompting tips for more productivity”. It means:
provide context. Don't just ask the question, provide the framework. Who is the target group? What is the intended use? Which perspective is relevant? The more context, the more parameters are activated, the more differentiated the answer.
Allow complexity. Many teams simplify their prompts because they think the model „understands” simpler instructions better. The opposite is the case. Simple instructions create simple processing paths. If you want differentiated results, you have to ask differentiated questions.
Dimension tasks appropriately. Not every task needs the full orchestra. A quick summary is a legitimate use. But if you only ever summarize, you are giving away 90 % of the paid capacity.
Experimentation instead of standardization. Many companies create prompt templates and distribute them to all departments. This sounds efficient, but it creates precisely the uniformity that relegates the orchestra to a solo. Instead: Encourage teams to test the boundaries. Try out different contexts. Observe where the response quality jumps.
Evaluate results, not just accept them. Perhaps the most important point is that most teams take the first answer and continue working with it. No questioning, no comparison, no iteration. But it is precisely in iteration that the leverage lies. If you tell the model „This is too superficial, go deeper into aspect X” or „Now argue this from the perspective of a skeptic”, you activate new layers of processing with every step. Every round of feedback is a new impulse for the orchestra. It's not the first answer that counts. It's the third one that counts.
The score decides
The metaphor of the AI orchestra is not perfect. No language model literally has musicians waiting to play. But the basic idea is right: The quality of the input determines the quality of the output - not linearly, but structurally. A rich context activates processing paths that lie fallow with a narrow prompt.
This is not a question of prompting. It is a question of understanding.
Anyone who understands what a language model actually is - namely not an intelligent text generator, but a statistical system with enormous depth and breadth - asks different questions. Not more clever. Not trickier. But more appropriate.
And that's the real point: most companies are not using AI incorrectly. They just use it flatly. They have booked an orchestra and are playing scales.
The score is yours.
How does your AI concert currently sound, if you're being completely honest?
0 Comments