Triage on arrival means a message is read and ranked the moment it reaches your mailbox, rather than the next time somebody sits down and works through the pile. The judgment is instant. The interruption does not have to be, and separating those two is the whole of what this is worth.
- A batch ranking expires the moment it finishes. Everything that lands afterwards is unsorted, and that window is where the checking happens.
- Mail already gets handled the second it arrives, by you. The measured price of that is a reaction in seconds and a recovery in minutes.
- Checking less is measurably better for the person doing it, and instructions to check less do not survive a working week.
- Judging on arrival and interrupting on arrival are separate decisions. Only the first of them should be instant.
What does triage on arrival mean
Three things, and only the first is about speed.
A message is read and given its place as it lands, not at the next scheduled pass. The list in front of you updates itself, so what you are looking at is never a photograph of an earlier moment that you have to reload to correct. And because both of those hold, the claim “this is what needs you” is true at four in the afternoon as well as at eight in the morning.
What it is not is an alert. This is the misreading worth heading off early, because it inverts the benefit. A system that tells you the instant anything arrives has not triaged anything; it has just moved the pile onto your lock screen. Judging every message as it lands is precisely what makes it possible to interrupt you for almost none of them.
It is also not about mail moving faster. Delivery has been effectively instant for twenty years, and nobody’s problem is that messages take too long to travel. The latency that still costs a small firm money is the gap between a message arriving and anyone forming a view on it, which in most inboxes is measured in hours and is bridged by a person repeatedly looking.
The hole in a once-a-day ranking
Ranking the overnight pile before the owner opens the app is the largest single improvement available, and how email triage works sets out why. But a ranking made in one pass has an expiry date, and the expiry is the moment the pass finished.
Look at when mail actually arrives. Overnight accumulation is the visible half, because it is sitting there in a heap at eight o’clock. The rest lands across the working day, one message at a time, in the middle of everything else, at a hundred-plus messages a day. A morning ranking covers the half that is easy to see and says nothing about the half that arrives while you work.
So what do you do about the unranked half. You check. Not out of curiosity and not out of poor discipline, but because a ranked list that stopped being true at a quarter past eight has left you without an answer to the only question you actually have, which is whether there is something in there for you right now. Looking is the only instrument you have for answering it.
This is the mechanism underneath a caveat that choosing email triage software states flatly: no product stops you checking. That is true and worth keeping. What a live ranking changes is narrower and more useful. It does not remove the habit. It removes the reason, because the answer you were going to get by looking is already on the screen and stays correct without you.
What already happens when mail lands
Every arriving message is already triaged in real time. The system doing it is you, and somebody measured what that costs.
The cost of email interruption, published in the Journal of Systems and Information Technology in 2001 by Thomas Jackson, Ray Dawson and Darren Wilson, monitored fifteen employees of the Danwood Group in Lincoln over 28 working days, gathering more than 180 hours of screen recordings from clerks, programmers and managers going about ordinary work. Two findings have outlived the study.
The first is how fast people move. “The majority of emails, 70%, were reacted to within 6 seconds of them arriving and 85% were reacted to within 2 minutes of arriving.” The authors’ own comparison is the one that lands: that is quicker than letting a phone ring three times.
The second is what the move costs. “The time it takes the average employee to recover from an email interrupt and to return to their work at the same work rate at which they left it, is on average 64 seconds.”
Hold those two numbers next to each other and the arrangement looks strange. Six seconds is not long enough to rank anything. It is long enough to look, and looking is what happens: one message, out of order, out of context, judged against nothing, by the person in the building who can least afford the sixty-four seconds it takes to get back to what they were doing. The fastest triage in your firm is also the worst, and it runs on every message that arrives.
Fifteen people at one British company on Outlook 97 and Outlook 2000, in a year when the paper notes those employees did not receive excessive amounts of mail, is a narrow base, and the six seconds is a direction rather than a constant. The mechanism it describes has not narrowed at all. The researchers observed that most mail applications were set to poll every five minutes and that every employee had a new-mail icon, with 57% also getting a dialogue box. The arrival of a message was being broadcast to the one part of the system with the least capacity to absorb it. That is still the default arrangement, and the broadcast now goes to a phone as well.
Why checking less is not about discipline
The obvious remedy is to check less often, and it works. It is also, on its own, close to unfollowable.
In Checking email less frequently reduces stress, a 2015 paper in Computers in Human Behavior, Kostadin Kushlev and Elizabeth Dunn at the University of British Columbia gave 124 adults two different weeks. In one, they were asked to check email three times a day, with the mailbox closed the rest of the time and new-mail alerts switched off. In the other, they were asked to check as often as they could, keep the mailbox open and have every alert running. The order was randomised, and everyone did both.
The limited week produced significantly lower daily stress, and people felt less tense while doing something that mattered to them. The effects are modest in size and they are real.
One control makes the finding much more interesting than it first looks. Across the two weeks there was no significant difference in how many messages people received, or in how many they answered. The mail was the same and the work was the same. The only thing that changed was how often they went to look, and that alone moved how the week felt. Whatever the checking is costing, it is a separate charge from the email.
Then there is the part nobody quotes. Asked to check three times a day, participants managed 4.7. These were volunteers who had answered an advertisement about feeling overwhelmed by email, who only qualified if they had some flexibility in when they checked, and who had agreed to spend a fortnight being measured. Three was still not holdable. Their normal was about fifteen.
That is the strongest evidence there is that checking is not a character defect. It is a rational response to not knowing, and it is only going to stop when the not-knowing stops. What a calm inbox actually feels like is the same finding written as an ordinary Thursday.
Judged on arrival, delivered on your terms
So there are two clocks, and they have been wired together for no good reason.
The judgment clock should run as fast as the mail. A message that arrived two minutes ago should already have a place in the order, because the value of a ranking is that it is complete, and a ranking that excludes the last two hours is not.
The delivery clock is a different question entirely, and it belongs to you. Almost nothing that arrives is worth the interruption it currently gets.
There is a name for running them separately. In work presented at the tenth conference on user modelling in Edinburgh in 2005, Eric Horvitz, Johnson Apacible and Muru Subramani set out bounded deferral, which they describe as “a method aimed at reducing the disruptiveness of incoming messages and alerts in return for bounded delays in receiving information”, giving people “a means for balancing awareness about potentially urgent information with the cost of interruption”. Messages are held while you are busy and released when you are free or when a time limit is reached, with how much a message matters deciding what gets through sooner.
The trade there is not silence in exchange for risk. It is a small, bounded delay in exchange for the interruptions, and the reason it is a good trade is that the delay is bounded and the interruption is not.
But notice what bounded deferral needs in order to work at all. Something has to have formed a view on the message while you were busy, or there is nothing to prioritise the release by. Silencing everything is the crude version of the same idea, and it fails in the way you would expect: with every message made equally invisible, the not-knowing comes back within the hour, and you go and look. Judging on arrival is what makes selective delivery possible. Selective delivery is what makes the judging worth having.
What it does not fix
Five things, and the first is the one people are quietly hoping for.
It does not answer anybody faster. Ranking on arrival is entirely a fact about your side of the exchange. The client waiting on a decision waits exactly as long as before, and the honest risk runs the other way, since a message you know is safely ranked is an easy message to leave.
It cannot judge better than it can judge. Instant and correct are different properties, and a message that landed thirty seconds ago carries the least history of anything in the mailbox, which is exactly the case a ranking handles worst. Where that boundary sits is the subject of AI email triage.
It does not decide what deserves to reach you. Something has to be told what is worth an interruption, and that is a judgment about your business rather than a setting with a right answer. Getting it wrong in the generous direction rebuilds the problem exactly.
It does not reduce the volume. The same amount arrives. What changes is who meets it first, and whether meeting it costs a decision.
It will not cure the reflex. Six seconds is a habit built over years and the phone is still in your pocket. What changes is that looking stops paying, because the screen says what it said the last time you looked, and a habit that returns nothing fades on its own.
How Point triages as mail lands
Point reads and ranks everything that arrived overnight before you open it, into now, later and other. The part this page is about is what happens next: new mail is triaged as it lands, without a refresh. A message that arrives while you are reading takes its place in the order there and then, rather than waiting for a scheduled pass or for you to reload the window. So what the top of the feed says is a live statement rather than one that was true at eight o’clock.
What is allowed to reach you is separate, and yours. You decide what is worth an interruption and what can wait for the next briefing, so almost everything gets judged and almost nothing gets to reach across the desk. Where you do want to know at once, you can ask for that in plain words: a standing request to be told when a particular message arrives or a condition is met, watching everything that comes in and telling you once, at the moment it matters, rather than pinging you through the day.
Briefings pick up the remainder. The day arrives as one read, at times you choose rather than times somebody else chose, which is the bounded-deferral trade made concrete: everything that did not warrant an interruption is waiting together in one place rather than scattered across the hours it arrived in.
The judgment goes back to your actual mailbox too. Connect Gmail and Point’s three groupings appear as labels on the account itself, and anything landing from then on is read and ranked on arrival, so a message sorted at eleven in the morning is sorted in Gmail as well as in Point, not only in the window you happen to have open.
Around all of it sit the usual controls, and triage on arrival is what makes them worth setting. The judging runs at two in the morning and through the meeting, so the real question is what Point may do in the hours nobody is watching. That is a level for each kind of action, from suggest-only through review to handled outright, with review the default out of the box, which is why a message weighed at eleven is prepared rather than answered until you say otherwise. Judged on arrival and acted on unprompted are two different settings, and only the first is turned all the way up to start with. What Point did is written down in order in plain language, and a row in that record is one click from being taken back. Setting something aside works the same way live: you pick the hour it returns and it comes back on its own, which snoozing and setting mail aside goes through properly. The benefits page is the full inventory.
Common questions
What does it mean for email to be triaged on arrival?
That each message is read and given its position in the ranking at the moment it reaches the mailbox, rather than at the next time a person or a scheduled job goes through the pile. The practical difference is that the answer to “is there anything here for me” stays correct through the day instead of decaying from the moment the last pass finished.
Is that the same as getting a notification for every new message?
The opposite, and this is the distinction worth being clear about. Every message being judged is what makes it possible to interrupt you for hardly any of them. A product that alerts you on arrival has moved the sorting onto your lock screen rather than done it. Judging on arrival and reaching you on arrival are two settings, and only one of them should be turned all the way up.
How is this different from a morning digest?
A digest is a batch, and a good one. Its limit is that it describes the mailbox as it was when it ran, so everything arriving afterwards sits unranked until the next one. Live triage and a briefing are complements rather than alternatives: the ranking runs continuously, and the briefing is the moment you choose to meet the part of it that never needed to interrupt you.
Do I still need to check my inbox?
Less, and for a different reason. Most checking is not curiosity, it is a query about whether something is waiting, and the research above suggests people ask it something like fifteen times a working day. When the ranking is always current and you have said what is allowed to reach you, the query has already been answered. The habit takes a couple of weeks to fade, which is normal.
Does mail that arrives while I am in a meeting still get ranked?
Yes, and this is the case that makes the difference. Arrival-time triage does not depend on the app being open or on anyone being at a desk. The hour you were unavailable is the hour where a batch ranking is furthest out of date and where you would otherwise be checking a phone under the table.
The short version
Mail is already triaged the second it arrives, by a person who reacts in about six seconds and needs about a minute to get back to what they were doing. Moving that judgment off the person is the point, and doing it at the moment of arrival rather than once a morning is what makes the result trustworthy through the day: a ranking that expires at a quarter past eight leaves you with no way to know, and looking is the only instrument you have. Checking less genuinely helps, and asking people to check less genuinely does not work, because the checking is a rational answer to not knowing. Judge everything on arrival, interrupt on terms you set, and the looking stops paying for itself.
What counts as worth an interruption is the part that varies by trade. It reads one way in a law practice and another in a design studio, with six other versions written out.