The Enpath Standard
Seven questions before deciding.
A public test for whether an artificial-intelligence answer deserves to be acted on. It is free, requires no engagement with Enpath, and is applied to our own work first.
The test
The seven questions.
-
Why was this recommended?
The reasoning is visible, not just the conclusion.
-
What evidence supports it?
Sources are cited and can be opened.
-
What did it assume?
Assumptions are stated, not hidden.
-
What is still uncertain?
The unknowns are named, with how much they matter.
-
Who is accountable?
A named person, identified before the decision.
-
How can it be challenged?
There is a real way to question or correct it.
-
How can it get better?
What would change the answer, and how it improves.
The bright line
If those seven questions can’t be answered, the system isn’t ready for a decision — no matter how convincing it sounds.
Adoption
Three ways to use it.
- 01
Apply it to an answer
Before acting on an AI recommendation, check that it answers the seven questions. If it can’t, it isn’t ready for a decision.
- 02
Build it into procurement and policy
The seven questions work as verifiable requirements in public procurement, AI policies, and oversight protocols.
- 03
Demand it from your vendors
A system that cannot show its evidence, its assumptions, and its accountable person should not operate consequential decisions.
Version
Public and under revision.
The Standard was published early on purpose — version 0.1 — so that Enpath would be the first organization measured against it. Revisions are logged in public.
The methodological detail is in Methodology & integrity.