Use product context before generating the answer

Scope a task to one user’s screen

Agree on stable screen IDs and supported release and plan values with the product team. Define how the assistant behaves when those fields are absent.

1. Step 1

   Map the request

   Interpret “split a shipment” as Partial invoicing.

   Taxonomy

   Maintain user phrases and approved feature names together.

2. Step 2

   Read session context

   Receive the active screen, plan, and release.

   Product or customer data

   The application supplies authoritative session values and entitlement information.

3. Step 3

   Retrieve the task

   Find the task associated with the screen and feature.

   Content team

   Maintain self-contained steps, screen references, and a completion check.

4. Step 4

   Apply scope

   Keep the applicable release and plan branches.

   Taxonomy

   Use controlled applicability values that match the application’s context fields.

   System

   The retrieval integration must apply the filters and preserve their values.

5. Step 5

   Answer and observe

   The model presents the applicable steps.

   Model

   Generation must preserve command labels and task order. Missing context should produce a clarifying question.

This framework provides a practical baseline shaped by experience. Every organization operates under distinct technical and operational realities that require tailored design and governance. While you can adapt these measures independently, we frequently partner with teams to customize governance, integrations, and rollout. Performance outcomes will naturally vary.

Verify the result and handle exceptions

Walk through the answer in the actual interface for each supported context. Test an old release and an ineligible plan, and confirm the assistant does not combine their instructions.

When context is absent or the screen has no maintained task, ask a focused question or offer the supported help route. Send screen/topic mismatches to the documentation owner.