Oops! Something went wrong while submitting the form.
Sai reads your Gmail from a set window and sorts it into what needs a reply today, what can be archived, what is waiting on someone else, and anything with a deadline. Replies that need a straight answer are drafted in your tone and saved in Gmail. The summary arrives in Slack. Nothing is sent.
The recording is a real session. The sheet on the right is what it produced.
Sai opens each profile, pulls the signal, and writes the row, live, in a real browser.
Eight columns, sorted by score, with a source link behind every claim.
Gmail access and a Slack workspace for the summary. A time window to look back over.
A Slack summary with bucket counts, deadline items first, and one line per message explaining its bucket. Replies to straightforward messages drafted in your tone and saved in Gmail. Nothing sent.
About three and a half minutes. The recorded run triaged 51 emails from a 48-hour window in 26 steps and 3 minutes 25 seconds, including drafting replies and sending the Slack summary.
Run it every weekday at 8am over the previous 24 hours. The summary is in Slack before the inbox is opened.
Gmail's Priority Inbox ranks messages as they arrive. Superhuman and Shortwave rebuild the client around keyboard shortcuts and AI classification. Filters and rules move mail into labels before it is seen.
These are fast. Superhuman's keyboard-driven flow processes mail faster than the standard Gmail interface, and Priority Inbox costs nothing and runs continuously.
All of them produce the same kind of result: an inbox that is better organised than it was. Reading it still requires opening it.
The word describes sorting that happens before treatment begins — establishing what is urgent before deciding where effort goes.
Applied to email, the useful output at 8am is not a sorted inbox. It is a short statement of what the inbox contains: three messages need answering today, one has a Friday deadline, four are waiting on other people, the rest can be archived.
That statement can be read in thirty seconds, on a phone, before the day starts. A sorted inbox cannot — it has to be entered.
This task sends the statement somewhere else. From the recorded run:
"Done — triaged 51 emails from the last 48 hours. Summary sent to your Slack DM, drafts saved in Gmail."
The summary is in Slack. The drafts are in Gmail, waiting for the moment a reply is actually sent.
Three of the four buckets exist in some form elsewhere: needs a reply, can be archived, has a deadline.
The fourth does not. Waiting on someone else covers messages that have been handled — you replied, and the next move belongs to the other party.
Email clients classify by importance to you. A thread you have already answered scores low on every importance metric, so it sinks. It is not something to reply to and not something to archive; archiving it removes the only record that you are waiting.
This bucket records how long each item has been waiting, and marks anything over five days. A thread waiting two days and a thread waiting three weeks require different responses, and the age is the information that distinguishes them.
Replies are drafted for messages where the answer is a fact, a confirmation, or a scheduling response.
Messages requiring a decision that has not been made are excluded. A draft written for one of those produces filler — text that reads like a reply and commits to nothing, which takes longer to fix than to write from scratch.
Drafts match the tone and sign-off already visible in the account's sent mail. From the run's opening step:
"I'll start by loading the Gmail integration skill to get the full API, checking my memory for your usual tone/sign-off, and resolving the machine handle — all in parallel."
Everything is saved as a Gmail draft. Nothing is sent. The draft sits in the thread where a reply would go, so sending it is one action taken deliberately.
Run it each weekday at 8am over the previous 24 hours. The summary is in Slack before the inbox is opened for the first time.
Because each run reports its bucket counts, a shift in the numbers is visible across days — a rising "waiting on someone else" count over a week indicates threads accumulating without response, which no single day's summary shows.
