Useful automation needs clear inputs, visible progress, and an outcome someone can verify—not just a moving loading indicator.
Define the outcome first
Start with a task your team already understands: prepare a draft, organize incoming requests, or find relevant information. Describe what a good result looks like and who will use it. A narrow task with a clear finish is easier to evaluate than an assistant asked to handle everything.
Show what is happening
Separate receiving information, processing it, and taking an action. People should be able to tell whether a request was accepted, is still running, or needs attention. Motion can reinforce a state change, but a label and a persistent result must carry the meaning when animation is absent.
Put review at the right boundary
A generated suggestion and an action that affects someone else have different consequences. Decide which steps can run automatically and which need review. Show the proposed action and the information supporting it before asking someone to approve it. Make correction and cancellation part of the workflow.
Test the awkward cases
Try missing context, conflicting instructions, duplicate requests, and a service that does not respond. Ask whether someone can understand the failure and recover without starting over. Keep representative examples and review them as the workflow changes. A convincing demo is a starting point; repeatable, understandable results are the goal.