How Smart Summaries work

Smart Summaries turns a long back-and-forth into something you can read in seconds. Open a thread and Chiefy gives you the gist: the outcome, who’s waiting on what, and any deadlines or amounts worth noticing. This page explains what Chiefy looks at, what it produces, and when a summary is created.

What Chiefy reads

When you summarize a thread, Chiefy works from the latest version of the conversation — every visible message in the thread, in order. It uses the full message content, not just the inbox preview, so the summary reflects what was actually said.

A few things to know about scope:

• Your drafts are never included. Chiefy summarizes the messages that have been sent and received. Anything you’re still writing stays out of the summary.
• It focuses on the latest messages. The brief is weighted toward what’s most recent and relevant, so a long thread doesn’t bury the part you need.
• It’s grounded in the thread. Chiefy summarizes what’s there. It doesn’t invent actions or pull in information from outside the conversation.

What you get

A Smart Summary has two parts:

• Bullet points — a short, scannable brief of the latest messages (usually one to three bullets). These front-load the outcome and highlight things like deadlines, amounts, and key names so they’re easy to spot.
• A snippet — a single concise line that captures the thread at a glance, useful as a one-line label for the conversation.

Together they give you a fast read on a thread and a one-liner you can scan in a list.

Which AI models Chiefy uses

Smart Summaries runs on a large language model. Depending on configuration, that’s either OpenAI’s GPT (served through Microsoft Azure) or Anthropic’s Claude (served through AWS Bedrock). The specific model is set by Chiefy and can change over time as better models become available — you don’t choose or configure it.

Whichever model is used, Chiefy guides it with a consistent set of instructions so summaries stay accurate, concise, and in the same house style across threads.

The model can differ between an on-demand summary you trigger yourself and one Chiefy prepares in the background. Both follow the same instructions and produce the same kind of brief.

When summaries are created

Summaries happen in two ways:

• On demand. When you ask for a summary yourself — for example, by choosing to summarize a thread — Chiefy generates it right then and shows you the result. See Summarize a thread.
• In the background. When new email arrives, Chiefy can prepare a summary automatically so the brief is already waiting when you open the thread. This runs quietly in the background and doesn’t hold up your inbox.

Either way, the summary is saved to your account and tied to the thread it belongs to, so reopening the thread shows the brief without regenerating it. When the conversation gets a new reply, the summary can be refreshed to reflect it.

How summaries connect to replies

While Chiefy summarizes a thread, it also weighs how likely it is that the thread needs a reply from you. That signal is what lets Autodraft decide which threads are worth pre-drafting a response for — so the summary and the draft come from the same read of the conversation. You can learn more in Turn on Autodraft.

Smart Summaries only ever produces a summary. Suggesting or preparing a reply is a separate step handled by Smart Drafts, and nothing is ever sent for you.

Your data and privacy

Chiefy reads thread content to write the summary, and it stores the resulting brief — the bullet points and snippet — alongside your account so it’s there when you come back. Summaries are derived from the thread you’re viewing; Chiefy doesn’t summarize messages you haven’t connected it to.