Distill a Long Email Thread into Action Items

This workflow takes a sprawling email thread — the kind with twelve replies, three forwards, and a "per my last email" buried somewhere — and turns it into a crisp list of action items. It fetches the full Gmail thread, feeds every message in chronological order to OpenAI, and asks the model to separate signal from noise: what was actually decided, who owns each follow-up, and which dates or deadlines were committed to. The result is sent back as a clean digest email so the takeaways live right next to the original conversation.

Use it when a thread has gotten too long to safely skim and you need the "so what do I actually have to do" answer in seconds. Good fits: a vendor negotiation that spanned a week, a cross-team launch plan with shifting owners, a customer escalation thread before you reply, or a Monday-morning catch-up on a conversation that ran over the weekend. It is especially useful before forwarding a thread to someone new — they get the summary instead of scrolling through forty quoted replies.

A new message arriving on a watched Gmail label (or thread) fires the trigger. The workflow pulls every message in that thread and stitches them into a single chronological transcript with sender and timestamp on each turn. That transcript goes to OpenAI with a structured prompt that returns three buckets — Decisions made, Action items (each with an owner and a due date when one was stated), and Open questions still unresolved. The workflow formats those buckets into a readable HTML digest and sends it via Gmail as a reply on the same thread, so the action items stay attached to their source. Because it runs on Agent Hive's Sim engine, every step executes for real against your connected Gmail and OpenAI accounts — no mock nodes, no "configure later" placeholders.