Ten questions that show where AI-supported analysis will serve a decision well, and where its confidentiality, ambiguity or political weight calls for human operators.
This is not a referendum on AI, which handles structured analysis better every year. It measures something narrower: how much of this specific decision lives in places analysis cannot reach: context you cannot share, questions you cannot yet frame, and consequences someone must personally stand behind.
Last reviewed 3 July 2026 · Free and ungated
The decisions this calculator steers away from AI are exactly the ones worth putting to the Global Board. Selected senior operators return decision-specific challenge, confidentially, on the questions analysis cannot carry.
Bring independent operator perspectives into this decisionAnswer for the decision itself, not for your organisation's general attitude to AI.
On these answers the decision is well defined, the context is shareable and the consequences are recoverable. This is the territory where AI-supported analysis genuinely excels: structured options, documented patterns and no confidentiality constraint on what you can feed it.
The decision has a substantial analytical core that AI handles well, wrapped in dimensions it handles poorly: partial ambiguity, some political weight, context you would hesitate to paste into a prompt. Neither pure analysis nor pure judgement is sufficient on its own.
The weight of this decision sits exactly where AI-only analysis cannot reach: context too confidential to share with any external tool, ambiguity where framing the question is the real work, and political or reputational stakes that demand judgement someone can personally stand behind. That is not a limitation of any particular model; it is the nature of the decision.
AI-supported analysis keeps getting better at exactly what it is built for: structuring options, processing evidence, surfacing patterns across documented experience. The interesting question for a decision-maker is no longer whether to use it, but which parts of a given decision it can carry. Some decisions are almost entirely analysable; others carry confidentiality, ambiguity or political weight that no analysis, however capable, can hold on your behalf.
The ten questions probe three things analysis cannot reach from outside the organisation. First, access: whether the real context can even be shared with an external tool. Second, framing: whether the work is answering a clear question or discovering what the question is. Third, accountability: whether the consequences require a person with lived experience to stand behind the call. High scores on those dimensions do not make AI useless; they define what remains after it has done its part.
Each answer adds 0, 5 or 10 points; the total, normalised to 100, measures how much of the decision sits beyond pure analysis. Three bands turn that into a recommendation for how to staff the decision. The calculation itself is fixed arithmetic in your browser. Fittingly, no AI is involved and nothing is stored.
No. Most decisions score in the ranges where AI-supported analysis is either sufficient or valuable alongside human review. The calculator identifies the narrower set where confidentiality, ambiguity or accountability mean analysis alone cannot carry the weight.
Partly. A private deployment eases the data-sharing constraint, but the question is also about context that exists in nobody's documents: board dynamics, unwritten commitments, the history behind a relationship. That context reaches a human adviser in conversation and reaches no system at all.
That disagreement is information, not an error to resolve by authority. Run this calculator on the decision: a high score suggests the disagreement lives in factors the analysis could not see, which is exactly when an independent human perspective earns its keep.