Because five forces act at once: everyone arrives prepared, the average solution acts as an anchor, every statement is auditable live, you consult twice, and increasingly agents find you instead of people. Whoever sees only one underestimates the situation.
The shift is not a single phenomenon but five forces acting at once. They interlock, and it is precisely their interaction that constitutes the half-knowledge era.
The first force: everyone arrives prepared
First contact no longer begins at zero. The customer has researched, generated concepts, sometimes sent whole architectures and lengthy documents they do not themselves grasp. You start in the middle of someone else's edifice of thought, which you must first untangle before you can build.
The second force: the average solution as anchor
What the AI produces is usually the median of all solutions in the training corpus, solid, generic, context-blind. The customer anchors to it, this is how it's done, and every better, more specific solution you must now sell against that anchor. This is the anchoring effect from behavioral economics, except the anchor now comes from a machine that appears impartial.
The third force: live auditability
The customer used to have to believe you. Today they type your statement into the AI and get a second opinion instantly. Every offer, every justification, every price is verifiable in real time. This is not only risk but also opportunity: whoever argues transparently and AI-robustly gains trust immediately.
The fourth force: the double consulting burden
You have to deliver two services where one used to suffice. First, correct and refine the knowledge brought in, from half-knowledge to a viable foundation. Second, build the actual conception on top of it. The insidious part: the first step almost always happens before the contract, that is, unpaid.
The fifth force: agentic discoverability
Increasingly it is not humans via Google but AI agents that find and evaluate providers. Your visibility depends on whether the machines understand you and pass you on.
The field calls this the shift from SEO to GEO, from search-engine optimization to optimization for generative engines.
Verifiable citations, source references and statistics significantly increase a source's visibility in generative answers, in individual queries by over 40 percent, while keyword stuffing hurts (Aggarwal et al., "GEO: Generative Engine Optimization", KDD 2024).
Example: the quote for the website
The customer called, described their business in three sentences and asked for a quote. You asked questions, wrote a concept, named a price.
The customer sends a twelve-page concept an assistant wrote for them, with a technology recommendation and a budget range. Your quote is typed into the same AI the same evening and compared with the median.
The next customer sends no text but has their agent compare three providers. Whether you make the list is decided by the machine from what it can verify about you.
Ask first: "What did the AI tell you about this?" Whoever begins with that question does not adopt the other's edifice of thought but understands it. And make the first step, refining what was brought in, visible and paid: that is what the gate in chapter 12 is for.
- Next in the Guide: chapter 4, The psychology of half-knowledge
- Course: Your agent, your rules (coming soon)
- Event: AI After Work, KI zum Feierabend, 23.09.2026, Ulm