Organizational Schema Theory · decision layer

Bring your AI the decision, not a search

You and your team already use AI to think through brand and operating decisions. Spectral Consult points that AI at exactly which research to read and which instrument to measure with — so the answer rests on measured work instead of a confident guess. When the evidence isn't strong enough to call it, it says so, instead of bluffing. It is a structured way to organize a decision over an open corpus — its sibling, the Brand Spectrometer, is the way to measure.

Use it in three steps

1. Tell your AI the decision you're facing — in plain words ("we're losing younger buyers," "should we acquire this brand?").
2. Point it at this page or at consult.orgschema.com/consult.json — the machine-readable index it can fetch.
3. It reads the right research and runs the instrument on your own material — and tells you where the evidence runs out.

Your data stays with you. The lens (the corpus + method + instruments) is ours, published, and kept current in the cloud — your AI reads it live, so it is always the latest science. Your case and the reasoning are yours.

Which decision are you making?

Each decision routes to the research that frames it, the concepts to share with your AI, and a runnable instrument.

The unbundled think-tank

A think-tank is two things in one building: the people who think, and the body of knowledge they think with. Spectral Consult splits them.

• We are the tank — a calibrated body of knowledge, method, and measuring instruments. Open, versioned, kept current in the cloud, so your AI always reads the latest science.
• You bring the think — your own AI, local or cloud, does the reasoning.
• Your data stays with you — your case and the decision you face never leave your side.

The lens is ours and published; the thinking and the data are yours. See how a decision is routed for the full three-part breakdown and the L0–L3 detail dial.

What it is — and isn't

• Bring a real decision, not a topic. The sharper the decision, the more precisely it routes.
• Read only what the decision needs. It points you at a few papers, not a reading list.
• Get a measurement, not an opinion. For anything quantitative, it runs the instrument on your own material rather than guessing a number.
• Trust the abstention. When it says "can't resolve this from the evidence," that is a real, useful answer — not a dead end. See a worked case where the abstention is the most useful line in the report.

For your AI assistant and developers

The surface is machine-first. An agent fetches one manifest and learns what to read and which instrument to run for each decision; every endpoint is CORS-enabled and open.

• /consult.json — the agent-facing manifest of decisions.
• /specs/<decision>.md — the full routing spec per decision.
• meter.spectralbranding.com — the Brand Spectrometer, the measurement instrument.
• /llms.txt — the corpus index for agents.

Read the work

The four typed verdicts and the noise-floor rule behind the routing are open — read the paper and run the code yourself.

Code open-source under MIT; theory published under CC BY 4.0. A projection of the open spectral-branding corpus — github.com/spectralbranding.