What Decision Grade means
A use-bound determination, the limits it preserves, and the different standard for work produced after the review.
The definition. Decision Grade is earned through a defined full review for one declared use. It is not another name for important work, a permanent document label, or a guarantee of correctness.
The term is useful only if it stays narrow.
A document is not Decision Grade in the abstract. It may earn that disposition for a specific board decision, investment review, filing, or approval—and fail a different use that asks more of the same material.
The declared use comes first
The declared use names the decision, action, audience, or reliance the document is intended to support.
That use determines what matters. A missing market boundary may be tolerable in an early working brief and decisive in an acquisition memo. A scenario may be useful for discussion and inadequate as the sole basis for a capital commitment.
Without a declared use, “quality” becomes a general impression. With a declared use, the process can ask a concrete question: is the unresolved work acceptable for this reliance?
What the disposition says
It says
- The required full review completed
- The result is bound to a named document version
- The use was declared
- Material findings and limits were preserved
- The recorded determination permits that declared use
It does not say
- Every statement is true in the outside world
- No future event can invalidate the document
- The document is fit for every audience or decision
- A human no longer owns the decision
- Later work inherits the determination
Decision Grade is earned through a full review. Completion alone does not award it. A completed review can still produce a hold, a refusal, a limited result, or named work that must happen first.
Triage and full review do different jobs
| Process | Purpose | What it can do | What it cannot do |
|---|---|---|---|
| Triage | Fast review while work is changing | Indicate possible issues in either reliance lane | Clear Decision-lane reliance |
| Full review | Deeper pre-reliance process | Support a recorded use decision after required work completes | Guarantee clearance or universal safety |
Triage is not a low-quality substitute, and full review is not a seal. They answer different workflow needs.
The integrity process
A serious pre-reliance check should make the following visible.
Source fidelity
Establish what entered the process, what could be read faithfully, and which material uncertainty must travel forward. Fidelity to a source is separate from whether the source is true.
Typed checks
Separate determinable results—such as arithmetic, schema, completion, and exact-reference consistency—from model judgment about support, assumptions, omissions, and reasoning.
Source binding
Tie every reader-facing finding to the author's actual material. A true observation that cannot be located and explained is not yet a useful finding.
Challenge and validation
Give material findings a separate attempt to fail. Preserve disagreement. Confirm that the required work completed and that no malformed or partial result passed as complete.
Disposition and record
Return the allowed action, any condition or hold, remaining outside-research needs, and the evidence needed to understand why.
Independence is a work design
Different models can share training data, habits, and blind spots. Provider count is therefore not proof of independence.
Useful separation comes from how the work is arranged: fresh source-bound reads, distinct jobs, controlled context, deterministic checks, governed challenge, validation, preserved disagreement, and a rule that no stage clears itself.
Different providers may strengthen that separation in some configurations. They do not create it automatically.
After review: decision readiness, not label inheritance
A review checks a bounded source artifact. Work after it can add synthesis, outside research, scenarios, backcasting, implications, and recommendations.
Those steps are valuable because decisions often require more than one document. They also increase epistemic distance. New reasoning introduces new assumptions and new ways for specificity to outrun support.
The Decision Grade determination therefore does not transfer to newly authored work.
The right standard is decision readiness:
- stronger source material;
- explicit assumptions and support states;
- visible reasoning dependencies;
- separate challenge;
- preserved uncertainty and disagreement;
- conditional scenarios rather than disguised forecasts; and
- a clear statement of what would change the answer.
This should produce a much stronger basis for judgment than an ordinary prompt. The size and consistency of that advantage remain to be measured.
Scenario planning and backcasting
Future-facing structures should be treated as conditional decision aids.
A responsible scenario states its starting facts, assumptions, branch conditions, horizon, disconfirming signals, and implications if the branch occurs. A responsible backcast separates the desired future from the evidence that each backward step is necessary and feasible.
Neither becomes verified merely because its sources were checked. The farther the work travels from evidence, the more important it is to show the route.
The operating test
The plain-language standard is:
Can a responsible reader see what was checked, what the material supports, where judgment entered, what remains unresolved, and why this use is allowed or held?
If the answer is no, the result may still be helpful. It has not earned reliance.