Case study / AI Automation
A helpful assistant behind every gathering.
A proposed automation workflow that turns incoming requests into organised drafts, with people in control of the messages members actually receive.
- Role
- Service design & automation UX
- Timeline
- 4 weeks · example scope
- Scope
- Workflow design · Review queue · Recovery states
Illustrative draft Sample story, scope, and concept screens. These are placeholders, not verified project results.
01 / The challenge
The same questions. A human answer.
In this sample scenario, community hosts repeatedly answer questions about upcoming events, joining a chapter, and updating an RSVP. The information is often straightforward, but the back-and-forth interrupts the work of actually hosting.
The design challenge was to reduce repetitive preparation while preserving a personal, accurate response. An assistant should know when it has enough context to help and when a request belongs with a person.
02 / The approach
Automate the preparation. Keep the judgment human.
The proposed system separates understanding, drafting, and sending, so an uncertain answer never quietly becomes a message to a member.
Start from approved information
Drafts use the chapter’s published event details and a reviewed FAQ. The review panel shows the source alongside the suggested response so a host can check the answer.
Route uncertainty to a person
Missing event information, ambiguous requests, and sensitive issues go to a manual queue. The interface explains what is missing instead of presenting a confident guess.
Make recovery part of the flow
A send failure stays visible with its status and retry history. Duplicate requests are grouped, and approval is required before the proposed workflow sends a reply.
03 / A closer look
04 / The reflection
Less repetition, with room for care.
The deliverable is a workflow concept and review experience, not a connected automation service. A pilot would first run in draft-only mode and track factual accuracy, host edits, and requests correctly escalated to a person.