Hotel guests ask open questions that the team cannot list as topics in advance, and a prose answer drawn from the property guide is an acceptable reply. Which orchestration approach fits?
Select an answer to reveal the explanation.
Short Explanation
Guests ask questions no card can list in advance, and a clear paragraph from the property guide is enough. You use generative orchestration so an unlisted question can be answered from that guide. A trigger phrase for every possible question is a list the hotel cannot finish.
Full Explanation
Generative orchestration fits when the questions cannot be listed as topics and a prose answer from approved knowledge is acceptable. The property guide is that knowledge, and guests will ask things the team did not anticipate, so the agent should answer from the guide rather than wait for a topic nobody wrote ahead of time. A growing list of trigger phrases tries to name every open question in advance, which is the list the hotel cannot finish. A closed language project that requires every guest intent to be enumerated before an answer is allowed blocks the long tail the scenario accepts as prose. A task agent that books or changes a room on any question takes an action the guest did not request, and a question about the pool is not a reservation change. The guide still has to be the source the orchestration reads, or the prose will not stay on property facts. Exam caveat: generative orchestration is a weak fit when the business requires a specific topic and a captured entity. Here a prose answer is the accepted result. Operational check: ask a guest question that matches no topic and confirm the reply is prose drawn from the property guide, with no room change made.