An AI decision-architecture diagnostic
Once AI retrieves information, interprets policy, recommends decisions, uses tools, or acts inside your operations, the question stops being how good is the model and becomes whether the system around it is organized to decide well under real conditions. Eight questions will show you where yours is exposed.
Most organizations still evaluate AI by asking whether it produces good answers. That is a component metric. The unit that matters is the operational decision — and every consequential decision depends on what the system knows, which logic governs it, what it is allowed to do, who can stop it, and what evidence remains afterward.
This diagnostic runs against a system you already have, a pilot that has stalled, or a build you are about to greenlight. It is not a technology audit; it asks nothing about which model or framework you chose. For each of eight dimensions you rate two things.
Answer for one system. Both ratings are needed before a dimension appears on your map.
Each dimension sits where its control meets its consequence. The top-left corner is where systems quietly injure people and institutions.
Where you're exposed, the remedy is almost never a larger model. It is better representation, clearer logic, explicit decision rights, an evidence trail, or real human authority — all of it within your control, none of it waiting on the next frontier release. The primer sets out the full framework the eight dimensions are drawn from.