Stating an eligibility rule without deciding a case

This is the whole discipline of the intent and it is one sentence: the assistant may say what the rule is, and may not say whether this visitor satisfies it. Everything anybody types about themselves is unverified, and an eligibility answer given on the strength of a self description is a decision made on no evidence.

The wording matters more than usual. Somebody who has just explained their circumstances and gets a bare rule back can feel ignored. The reply that works states the rule, says clearly that it is the rule rather than a decision about them, and names who checks it and when. That is not a refusal. It is the information they actually need in order to know where they stand.

It also protects the visitor. A yes from a widget is not a right of entry, and somebody who travels on the strength of one and is stopped at the door has been failed by an answer that sounded generous.

The follow up is always some version of can you make an exception. Naming the source of the rule answers it before it is asked, and does so without an argument. A statutory limit and a house preference invite completely different conversations, and the visitor can tell which one they are in.

It also stops the assistant sounding arbitrary. A rule with a reason attached reads as a business that has thought about it. A bare number reads as an obstacle to be negotiated with whoever answers next.

Half the value in this intent is logistical rather than legal. Somebody planning to come needs to know what to bring, whether a photograph of a document counts, and whether the check happens before they travel or after they arrive.

Publish that and a whole class of wasted journeys stops. It is also the part of the answer that is unambiguously safe to give in full, because it is identical for everybody who walks through the door.

Some questions are deliberately close to the line and some are close to it by accident. Either way the assistant should not reason its way to a verdict from an adjacent rule, and it must never combine two published conditions into a conclusion about a case that neither of them covers.

Configure this cluster on the cautious side, so a weak match returns your fallback rather than the nearest confident sounding paragraph. The cost of refusing is one message to a person. The cost of a wrong yes can be somebody excluded on arrival, an insurance position weakened, or a rule broken in your name.

That covers nearly every real message in the intent, which is why the route out has to be warm rather than reluctant. The assistant gives the rule, says a person can confirm the case, and takes the details.

The handover earns its place here because the person receiving it can often answer in one line, and because a case that turns out to be fine is a customer who nearly did not come.

Restricted goods, anything involving children, anything where a licence or a medical condition is in play. These carry consequences beyond a lost booking, and in some of them a wrong answer is an offence rather than a mistake. Route them to somebody accountable and keep the assistant to describing what is published.

Somebody explains their situation, the assistant finds the closest published condition, and answers in a way that reads as approval. They plan around it. Somebody travels, or takes a day off work, or brings a child, and at the point where the rule is actually applied the answer is no.

The whole cost of that lands on the person who was told and none of it on the system that told them. What makes it worse than an ordinary wrong answer is that they will reasonably say they checked, and they did. Keep the material to rules rather than to reassurance, name the checker and the checking moment in every answer, and run this intent cautiously enough that a case it cannot cleanly match becomes a handover instead of a guess.