Teach your human method to AI so it carries your judgment
- Tom Hansen
- Aug 31
- 4 min read
Leaders hit the same wall. A large model can answer, summarise and draft, yet the output feels thin where it matters. The model holds knowledge about the world, while your best work rests on knowledge about you. It cannot see the inner structure that guides your choices when stakes, context and time pressure shift. This article lays out how to surface that structure and teach a model to follow it with integrity.
Above is the 20 minutes podcast explaining why, how, what. And below is the prompt to do it for you.
Why capable models still fall short on expert turf
Across expert domains leaders report the same experience. When they judge outputs inside their own field, they often find the result weaker than what they would produce themselves, even after careful prompting. The issue is not a lack of information. It is a lack of access to a personal way of thinking that has been built over years of work, education and lived practice.
This is the human method, the distinctive approach people try to make a model imitate. It rarely succeeds while the method remains unspoken. Many are only dimly aware of it, which makes transfer hard until the method is brought into view.
What a human method really is
A human method is the implicit epistemic structure an expert uses to solve problems inside a familiar domain. It is a living system formed by experience, intuition, a rhythm of inquiry and unspoken logic. It carries the small distinctions that steer judgment and the language patterns that help meaning land in a team. Because much of it sits below the line of conscious description, it is seldom transferable to AI without an interpretive effort that makes the implicit stable enough to teach.
The traits that make your thinking teachable
Situational judgment sets weight on signals in response to the room you are in, for example a board review versus a customer escalation. Epistemic sensitivity notices when something is off even before the reason is clear. Reflective loops interrupt your own thread to test the frame and switch vantage point before commitment. Style acts as a carrier of meaning because sentence length, qualifiers and cadence guide the reader toward what is evidence and what is interpretation. Habit shaped cognitive economy lets you move fast because countless micro decisions have become second nature. Together these traits define the brief for what a model must learn if it is to carry your way of thinking.
From hidden method to teachable architecture
Begin with archaeological mapping. Observe your own best work and capture recurring moves. Write down the question you always ask before you greenlight a decision, the early warning sign that makes you slow down, and the linguistic choices that lift signal and remove fluff. Design an epistemic container that encodes friction and gates where drift tends to begin. Friction is a deliberate pause that forces a check when momentum would otherwise glide. Gates are decision points that require a test, for example a weakest plausible counter case or a restatement for the person who must act. Finish with linguistic synthesis so the model uses sentence patterns and vocabulary that carry your reasoning with the same pressure and clarity as your own writing. This interpretive work turns a personal method into a shared asset.
A Danish B2B scene that shows the point
Picture a commercial leader in an industrial firm who runs a weekly pricing and pipeline review. The ritual is stable, the inputs are repeatable and the downstream effects are measurable. The leader studies recent quarters and surfaces three fingerprints. Any recommendation must show the trade off between margin integrity and strategic account momentum. A red flag appears when a forecast argument uses adjectives where a number should sit. Before a bet is approved, the text must name the first reason it could fail within the next sixty days. These fingerprints become the spine of the container. The friction is a short pause before final wording. The model must surface the weakest counter case, restate the action for the account owner who will carry it, and isolate the number that resolves the claim.
The linguistic synthesis keeps the opening sentence short, puts the decision at the top and
moves evidence before narrative. After a month the review feels different. Inputs are shorter and clearer, the leader spends less time rescuing weak drafts, and decisions age better because the same method keeps running in the background.
How to use the prompt
The prompt below is both in English and Danish.
This prompt helps you make your own method visible and usable with a model. It guides you through the process.
It only works if you have memory features toggled. All you have to do is copy-paste it and say ok to continue with the next layer.
If you don’t have memory toggled, enclose the document and ask your model to convert this prompt and its layers into questions, and set aside 1 hour.
With memory toggled is takes less than 5 minutes.
Feel free to send me questions or comments.
