Your senior people cannot tell you how they decide. That is not evasion, it is Polanyi’s paradox, and a structured interview is the only reliable way around it. This worksheet is the interview.
The script runs in five sections: the decision being studied, the inputs the expert actually uses, the heuristics they apply, where those heuristics break, and the final check before they commit. Twenty-six questions, each with room to write. The questions are deliberately concrete. “Describe the last time you made this decision” surfaces things that “how do you make this decision” never will.
The interviewer instructions matter as much as the questions. Run it one-to-one, never as a group, because group dynamics suppress exactly the knowledge you came for. Start from a specific recent case, not a general description. And when the expert says “I just know,” slow down instead of moving on. That sentence is the signal that you have found the tacit layer.
A synthesis worksheet follows the interview. You draft the decision rule in one or two sentences, rank the top five inputs, state three to five heuristics as IF/THEN pairs, list the edge cases where the rule stops working, and score your own confidence on completeness, accuracy, and generalizability. Then you test the rule against twenty historical cases before anyone writes a prompt.
Use it before the automation scoping conversation, not after. The output is the substrate an agent needs, and it is also an honest map of which parts of the work should stay human.