​

Outputs

Outputs are the concrete work products a session asks for. They come in two kinds: In decision tasks, the output is the whole task: the operator reviews the inputs and produces one judgment. In conversation scenarios and journey simulations, outputs are produced alongside the dialogue, the way a case manager writes file notes during and after a call.

​

Inline or from the library

You can define an output in two places:

• Inline on a single scenario, when the output is specific to that situation
• Linked from your organization’s library, when the same output definition applies across many scenarios

A referral letter format that every scenario in your claims domain shares belongs in the library. Define it once, link it everywhere, and an update to the definition reaches every scenario that links it.

​

Simple and advanced modes

Each output definition uses one of two modes: Simple mode covers most cases. A summary note is free text. An approve-or-deny decision is a predefined option list. Reach for advanced mode when the output feeds a downstream system or has a fixed structure, such as a referral that must always include a recipient, a reason, and an urgency level.

​

Exemplars

An exemplar is a model output attached to a definition: a complete example of what a good artifact or decision looks like for this scenario. Exemplars do two jobs:

1. They make your standard concrete. “Write a good summary note” is vague; a model note is not.
2. They sharpen scoring, because the criteria for a good output are grounded in an example of one.

​

Next steps