Email triage is the pass you make over unread mail to decide, message by message, what needs you now, what can wait, and what needs nothing at all. It is a ranking job rather than a reading job. Done by hand it holds at forty messages a day, and it breaks somewhere past a hundred.
- Triage asks three things of every message: does this need me, does it need me now, and how will it come back if it does not.
- The cost is the deciding, not the reading. At a hundred-plus messages a day, the deciding is the part that stops fitting.
- Hand triage fails in a documented pattern: arrival order first, easy mail cleared first, important mail left for a second pass.
- A triage system finishes the ranking before you sit down, so your decisions start from a list that is already sorted.
What is email triage?
The word comes from medicine, where triage means sorting the people who have just arrived by how badly they need attention, so the worst case is seen first rather than the earliest arrival. Email borrowed it for the same reason. Mail turns up in the order it was sent, and that order has nothing to do with the order it should be dealt with.
The research defines it plainly. Exploring Email Triage, a 2019 study by Bahareh Sarrafzadeh, Ahmed Hassan Awadallah and Milad Shokouhi for the ACM conference on human information interaction and retrieval, calls it “the process of going through unhandled email and deciding what to do with them”. Its evidence is fifteen interviews and 91 survey responses gathered inside one large US technology company, all of them Outlook users, which is worth holding on to when the numbers below get specific. The same paper draws the distinction that matters. Search means looking for one thing. In triage, “the user is making a decision about every email as she comes across it”.
Every email. That is the problem in two words.
Worth saying what triage is not. Not filing, which decides where a message lives rather than whether it needs you. Not inbox zero, a target for a count rather than a judgment about a message. And not replying: triage ends the moment the decision is made.
What does email triage actually decide?
Three decisions, in order, on every message.
- Does this need me at all? Mostly it does not. In Understanding Email Use, a 2005 Carnegie Mellon study, 124 people on a university-wide list rated five messages each from their own inboxes, 581 in all, and “a striking 64% did not require any reply”. You pay full price for deciding on all of it, and nearly two in three come back as nothing.
- Does it need me now, or later? This turns on two separate things: how much a message matters, and how soon. Treating them as one axis is the usual reason the message that will cost you most sits below the one shouting loudest, which important versus urgent takes apart properly.
- If later, how does it come back? The question people skip, and where the whole thing leaks. In the 2019 study only 18% needed no method for returning to deferred mail. Everyone else improvised: 32% marked messages unread, 18% flagged them, 2% moved them to a folder, 30% ran some mixture of tabs, reminders and memory.
Deferral is not an edge case, it is most of what triage produces. Three quarters of those surveyed deferred at least one message a day, and three quarters of deferred messages carried a task. The surprise is which ones: more than 75% needed under half an hour of work. People put off quick things, because breaking off costs more than the tracking will, right until the tracking fails.
How do you actually triage an inbox?
The procedure is short. Doing it well is mostly a matter of refusing four things the inbox invites you to do.
- Look at the whole arrival before you touch any of it. Ranking is a comparison, and you cannot compare a list you are reading one item at a time from the top. Take in everything that came in, then decide. A pass that starts with a decision on message one has already given up the ranking.
- Order by what it costs to ignore, not by when it landed. Arrival order is the order the inbox hands you and the only one guaranteed to bear no relation to what matters. Nothing about a timestamp tells you the note underneath it is the one that will cost you a client.
- Make one decision per message, and only that decision. Now, later, or nothing. Not a reply, not a folder. The moment you answer something you have stopped triaging and started working, and the rest of the list stays unranked behind you, which is how a pass that felt productive ends with the important mail still unread.
- Give every “later” a date, not a marker. An unread mark and a flag are both notes to your future self that say only “something here”, with no time attached and no way to tell two weeks-away task from a five o’clock one. Deferral is most of what triage produces, so the place you put things is not a detail of the method, it is the method.
- Finish the pass before you act on any of it. The ranking is worth having only if it is complete. Half a ranking is just the top of your inbox with extra steps.
Run that way, hand triage is a genuinely good system, and for a couple of decades it was the only one available. It is also the version almost nobody manages, and the reason has nothing to do with discipline.
Why does triaging by hand stop working?
Start with the budget. Microsoft’s June 2025 report Breaking down the infinite workday, built from global Microsoft 365 telemetry and a survey of 31,000 knowledge workers across 31 markets, found that “the average worker receives 117 emails daily”, “most of them skimmed in under 60 seconds”, and that employees using Microsoft 365 are “interrupted every 2 minutes by a meeting, email, or notification”. Under a minute of judgment per message, spent in fragments.
What that budget does to a triage pass was measured elsewhere. The figures that follow come from the 2019 study above rather than from Microsoft’s telemetry, so they describe 91 people inside one company, and they line up in a specific pattern.
The order is wrong from the start. More users “tend to triage emails sequentially than by priority (48% vs 41%)”, which is to say they work down the list in the order the inbox handed it to them.
The first pass then spends itself on the wrong mail. Users “tend to handle emails that are easy to delete or archive first (corresponding to 46% of all responses), followed by emails that are important to one’s work (28%)”. So the pass you reliably finish is the one clearing what mattered least.
And the pass that would have caught the rest is optional. Slightly over half of those respondents used more than one pass, while 46% ran a single pass only, on a day interrupted every two minutes since.
That is not carelessness. It is what a list showing only sender and time does to anyone asked to rank a hundred and seventeen things from it. Why important mail slips past follows the cost over a week.
What does a triage system do instead?
The change is not that deciding gets faster. It is that the ranking is finished before you arrive. Laid side by side, the same five steps come out differently at almost every line.
| Triaging by hand | A triage system | |
|---|---|---|
| Who ranks it | You, before you are allowed to start work | Done before you open the inbox |
| Coverage | As far as you got before the interruption | Everything that arrived, complete |
| What it ranks on | Sender and time, which is all a flat list shows you | Importance and urgency, weighed as two things |
| Cost per message | One decision, on every message, every day | A decision only on the few that need one |
| What “later” means | An unread mark, a flag, and your memory | A date you chose, which comes back on it |
| Order you work in | Whatever arrived most recently | What it costs to ignore |
| Where it breaks | Somewhere past a hundred messages a day | On judgment calls, which is why it should hand those back |
Four things separate a system that holds from a habit that does not.
- It reads all of it, not the top of it. A person triages until they are interrupted. A system covers everything that came in overnight, so the ranking is complete rather than as far as you got.
- It weighs importance and urgency together. Something that matters enormously and is not due for two weeks is not the same as something trivial due at five. Both axes, on every message, is the judgment a rule cannot make and a flag cannot record.
- It leaves few enough buckets to act on. Three is about what you can take in at a glance: needs me now, needs me later, needs nothing. Seven folders is filing again, which is the job you were putting down.
- It treats later as a real place with a time on it. Since deferral is most of the output, what you set aside has to return on a date you chose, rather than sit behind an unread mark hoping to catch your eye.
One more is easy to miss: it has to be reversible and on the record. If you cannot see what was decided and put it back, you have not handed off the sorting, only lost sight of it.
That is a specification rather than a shortlist. Working out which products actually meet it is a separate job, and the guide to choosing email triage software does it: what actually gets sold under that name, how to separate the genuine article from a rules engine wearing the label, and how to run a trial on your own mail before you commit.
Three habits hold whatever you run it on. Triage at set moments rather than continuously, because a few passes at times you choose cost far less than answering the same question every two minutes. Write down who always matters, since a named list catches the people you already know about, and VIP mail that never gets buried covers how to pick them. And judge the result by what reached you, not by the count: good triage means nothing important sat unseen, which is not the same as an empty inbox. Inbox zero versus a calm inbox is the long version of why those differ.
Where does AI come in?
Late, and as the mechanism rather than the point. The method above has been sound for twenty years. What was missing was anything able to read a hundred and seventeen messages before you sat down and form a view on each, which is the part a person cannot do and a model can.
Point is built to run it. Everything that arrived overnight is ranked before you open the app, on both axes, so what needs you today leads the feed instead of whatever landed last. Set something aside and you choose the hour it comes back. What Point may do on its own is set per kind of task and starts on review, and there is a plain-language log of what it did, with a way to reverse it. Connect Gmail and the same three groups appear as labels on the account itself, so the ranking travels with the mail. The benefits page has the full inventory, and what an AI email client is covers the category.
Whether a model’s judgment is good enough to rely on is a fair and separate question, and the answer depends on what you are asking it to judge. AI email triage works through what a model can and cannot decide, and how to test one against your own mail before you lean on it.
Common questions
Is email triage the same as inbox zero?
No. Inbox zero is a target for how many messages are left in the list. Triage is a judgment about each message, and it is finished whether the inbox ends up empty or not. You can triage well and still have two hundred messages sitting there, every one correctly judged not to need you.
How long should email triage take?
Long enough to have a view on everything that arrived, which at a hundred-plus messages a day is not achievable by hand in one sitting. That is why most people run several passes and never quite finish. When the ranking is already done, the pass becomes a read rather than a sort.
What is the difference between triage and email rules or filters?
A rule is written in advance and applies the same way forever. It can catch a sender or a subject line, but it cannot tell that this particular note from a regular client is the one that matters today. Rules handle the predictable. Triage handles the rest, which is where the risk lives.
Does email triage mean messages get deleted?
It should not. Deciding a message does not need you is not the same as destroying it. A triage system worth using moves low-priority mail out of your way while leaving it findable, and lets you pull anything back if the call was wrong.
Who should do the triage in a small firm?
Whoever gets blamed when something is missed, which is usually the owner. That is the reason to hand the ranking to something that never gets interrupted, and keep the deciding for yourself.
The short version
Triage is deciding, not reading, and at a hundred and seventeen messages a day the arithmetic does not close. Getting faster at it will not help, because speed was never the constraint. What helps is opening a list that has already been ranked, and spending your judgment on the handful of calls that actually needed it.
What a ranked list is worth depends on what lands in yours. The same weighing reads differently in an accounting or tax practice than in a PR and communications firm, and there are eight versions of the morning written out.