The idea
Tasks and deadlines were arriving through several channels. I wanted to bring that information together, sort out what mattered, and get a useful to-do list without having to keep checking each source.
Quinn connected email and calendar information to an assistant I could interact with by text message.
My contribution
I developed Quinn with Claude's assistance, directing the workflow, configuring integrations, and testing the results. Claude generated the application code; my work focused on what the assistant should do and whether it was useful in practice.
When the initial approach struggled to produce a useful task list, I added a more capable model step to help prioritize information, identify duplicates, and separate actionable items from email noise.
How the pieces fit together
Bring incoming email, attachments, calendar events, and stored tasks into one workflow.
Clean up duplicate information, identify deadlines, and compose a brief around what needs attention.
Deliver a concise SMS update and support follow-up questions through the same channel.
The implementation combines deterministic task-handling rules with model-generated wording. Its briefing code includes a plain-text fallback when model composition fails.
What happened in practice
I completed the SMS setup and received real text messages from Quinn. The combined workflow worked well enough to help organize the information coming in.
I eventually stopped using it because Brightspace and email were already communicating much of the same information. The extra channel became redundant in my situation.
What I learned
An automation can work technically and still add little value to an existing routine. The important question is whether it reduces the effort of deciding what to do next.
Quinn also showed me that collecting information is easier than deciding what deserves attention. Duplicate handling, context, and prioritization mattered more than simply connecting another source.
Why it matters for business analysis and enablement
This project involved understanding a workflow across several systems, deciding which information was actionable, and refining the output based on actual use.
It is the kind of thinking I would bring to AI enablement: start with a real need, test the result in context, and be willing to simplify or stop an automation when it duplicates a process that already works.