A typed decision model returns a choice and a confidence, not text. The method governs it like any other delegate, and the standards name the six places its authority sits.
A typed decision model takes a block of state and a set of questions with fixed answers, and returns a choice, a score, or a probability with a confidence margin. It writes no text and gives no reasons. The first widely adopted product in the class, TypeSafe AI's Jev, was released in September 2026 and marketed as hundreds of times faster and cheaper than a language model for routing, triage, tool gating, and approval checks.
The method needs no new primitive for it. The unit of governance is the consequential decision, whatever component made it. What changes is the price and the grain. When a judgment costs a fraction of a cent and a few hundred milliseconds, institutions put one at every branch, and authority moves into places that do not look like decisions.
The standards name six of those places: the threshold that decides between acting and asking, the content in the state the model reads, the queue a reviewer is left with, the hops that shape a decision without making it, the reasons nobody produced, and the rate at which decisions outrun recourse.
Canonical entry: Typed decision models and where their authority lives in the glossary.
A support platform puts a decision model in front of its refund agent. Refunds scoring above 0.8 run unattended and the rest go to a reviewer. The 0.8 lives in a configuration file, was tuned on the previous model version, and the reviewer only ever sees borderline cases. Under STD-08 §2.6 and §3.6, the 0.8 becomes a policy record pinned to the model version it was tuned on, and a declared share of the refunds that ran unattended is sampled into review, so a confident error has somewhere to be seen.
Implementation
Unique operational detail so this concept stands on its own in policy, procurement, and review workflows.
The number that sends a case to automatic execution or to a person is where authority is conferred. STD-08 §2.6 holds it as a policy record naming the labelled cases, model version, question schema, and option order it was set against. An unpinned model version means the trigger has already fired.
A gate that reads emails, pages, and tool outputs can be argued with by them. A forged note claiming the user already approved has moved a real gate across its threshold. STD-08 §1.5 reads authority only from grant and authorization records.
A queue fed only with low-confidence cases cannot catch a confident error. STD-08 §3.6 makes the routing threshold part of the intervention specification and samples decisions that skipped review back into it.
A classifier that ranks, filters, or routes what the decider sees is part of the decision in aggregate. A router that picks the model is a delegation over its whole range. STD-09 §1.5 and §1.6 enumerate both.
A model that gives no reasons invites a language model to write one afterwards. That text is an account of a decision it did not make. STD-02 §1.4 labels it as one and requires the decision object to say how the decision was produced.
At machine speed, a class can issue more errors in a day than its responder can answer in a year. STD-02 §8.6 bounds issue rate times error rate by answer capacity, and treats a class past that bound as uncontestable in practice.
Standard
The threshold as a policy record, content that cannot authorize, and review that samples what ran unattended.
Standard
Hops that shape a decision without making it, and routers that choose the delegate at run time.
Standard
Reasons that belong to the decision, and an issue rate bounded by the capacity to answer challenges.
Worked example
Validated records for a refund agent behind a decision model, with the field mapping from a typed model's response.
Evals
Seven draft cases test each of these places in a live deployment, from a forged approval in the state to issue rate against answer capacity.
A typed model cannot return an answer outside its schema. It can still return a valid, confident, wrong one. Schema validity is a property of the output format, and the method governs the decision the output becomes.
Not on its own. The vendor's own documentation describes confidence as a margin, not a probability that the answer is right, and calibration measured on someone else's workflows is not evidence about this deployment. Law III couples authority to evidence the institution holds. Confidence does not authorize.
Because the reasons would describe a plausible decision, not the one that was made. Recourse depends on the reason being something the affected person can dispute and that would change the outcome if wrong. An account written by another component meets neither test.
No. The method takes no position on which component decides. Typed output can make governance cheaper, because the choice, the distribution, and the confidence are exactly the fields a decision record needs. The risk is in treating a decision as plumbing because it was cheap.