What this pack runs.
A pilot demonstrates that a model can perform a task. Production requires that somebody can be accountable for the output when it is wrong, and only the first was ever tested. This pack sets out the five gates between the two, names which of the four postures a deployment is actually running, and prices the review break-even against the review share a measured error rate requires — then charges for the errors that survive review, which is the test most business cases never run.
The casebooks these cases come from.
The pack reads nine sourced decisions through the framework. They are drawn from these casebooks, each one a set of verified records on a single decision type.
| Casebook | Verified records |
|---|---|
| The Fall | 51 |
A verified record has passed a mechanical claim-versus-source check, a second model's read, and human approval. The count is what exists today, not a target.
Loading the rest of the sample…
16 files for this one decision.
The extract holds one case of the nine, and the number without the model that produces it. $499, one-time.
See the full pack →