Decisive AI · Vol. 5 · Appendix D

AI Intelligence Script — interrogate the machine

The Volume 1 Intelligence Script taught you to gather ground truth from people. This version teaches you to interrogate the machine: four copy-ready prompts for any consequential AI output, plus three questions only a human can answer.

Open the AI Intelligence Script

Decisive AI · Vol. 5 · Appendix D

AI Intelligence Script

The Volume 1 Intelligence Script taught you to gather ground truth from people. This version teaches you to interrogate the machine.

Zero-Retention Policy:Your data is processed and purged. We don't train on your decisions.

The Architect's Opening

When: Replacing "what should we do?"

Here is the situation: . Our objective is , with these weighted criteria: . Constraints: . Define success as: . Generate distinct options that fit, including at least one we probably haven't considered.

The Dissent Pass

When: Any recommendation you're about to act on.

Now argue against your own recommendation. What would have to be true for this to fail?

The Blind-Spot Probe

When: Any forecast or analysis.

What relevant factors can you NOT see from the data available to you? What has changed in the world that your information may not reflect?

The Sensitivity Test

When: Any ranking or scored comparison.

Which inputs or scores would have to change for your ranking to flip? How fragile is this conclusion?

The Three Questions the Machine Cannot Answer

Ask these of a human — ideally the one in the mirror:

  1. Should we?
  2. Who owns this outcome?
  3. Can I defend this to someone I respect?

No copy button — these are for you, not the model.

Built a full brief first? Continue from the Decision Design Brief →

Follow Decisive Leader on LinkedIn

About this tool

Four scripts, one discipline

The Architect's Opening replaces vague "what should we do?" queries. The Dissent Pass, Blind-Spot Probe, and Sensitivity Test stress-test whatever the model returns before you act.

Fill variables, copy, paste

Click any [VARIABLE] to fill it inline. Copy one script or all four into your own AI tool. Nothing leaves your browser.

Three questions for the mirror

Should we? Who owns this outcome? Can I defend this to someone I respect? The machine cannot answer those — the fifth card is addressed to you.

Browser-only · interrogation scripts

Copy these prompts into your own AI tool. Nothing is sent to our servers. Variable fill-ins stay in this browser tab until you refresh. Pair with the Decision Design Brief when you need the full human design pass first.

Was this tool useful?