Decision intelligence, not descriptive dashboards
From watching to acting — the move from a number on a screen to a cone a CRO can hedge.
A dashboard tells you what happened. A decision-intelligence system tells you what to do about what might happen. The first is a screen you watch; the second is a cone you hedge. The distance between them is the distance between a report and a decision — and a regulated firm cannot defend a report, only a decision.
Descriptive dashboards are the dominant form of "AI" in regulated firms, and they are the form that requires the least from the model. A dashboard shows you what already happened: last month's defaults, this quarter's exposure, the realized loss. It is a mirror, not a forecast. Decision intelligence is the move from watching the mirror to acting on a forecast — and the move changes what the system has to be. A dashboard can be wrong and no one cares; a forecast that drives a decision has to be defensible, and that is a different bar.
The cone is the deliverable
The decision-intelligence surface is a probability cone: a distribution over scenarios, with entropy, with modes, with calibrated tails. The CRO does not read a number off a screen; they read a shape and decide how much to hedge. The cone is the thing that turns "what happened" into "what to do" — because the shape tells you which contingency to plan for and with what weight. A dashboard cannot answer "how much should I hedge"; a cone can. That is the difference between a report and a decision.
Why the bar changes
A dashboard that is wrong is a nuisance. A forecast that drives a decision and is wrong is a liability — and a regulator will ask whether you knew it was wrong, could have known, and can show how you got it. That is the bar the dashboard never had to clear and the cone always does. The cone carries a per-output manifest (BR-842): which NodeVersions, which model, which agents, a replay anchor. The decision is defensible because the computation is reconstructable. A dashboard is not defensible because it was never a decision.
A dashboard can be wrong and no one cares; a forecast that drives a decision has to be defensible. The bar changes when the output becomes a decision, and the manifest is what clears it.
- A dashboard reflects; a cone projects. One you watch; the other you hedge.
- The cone carries the manifest — the decision is defensible because the computation is reconstructable.
- Calibration (Gate 1b) is what makes the cone trustworthy enough to act on; a dashboard has no such gate.
What we do not ship, and why
We do not ship a dashboard. We ship cones — distributions, with provenance, gated calibration, and a replay anchor per output. The dashboard is the form the regulated industry is comfortable with, because it asks nothing of the model and nothing of the firm. The cone asks both: the model has to be calibrated, and the firm has to decide. We are pre-revenue, building the POC; the cones are produced, the calibration that makes them act-worthy is gated (Gate 1b), and the manifest that makes them defensible is built. The move from dashboard to cone is the move from watching to acting — and only one of those is a decision a regulator will let you keep.
A dashboard is a report; a cone is a decision. The regulated firm cannot defend a report — it can only defend a decision, and the decision needs a cone and a manifest, not a screen.
— product note