It’s 8:50 and you’re finished.
The paper beside your keyboard says 9:20, in your own handwriting, from the first sorted morning of last week. So one of the two numbers has moved by half an hour. It’s an ordinary Wednesday with clients in it, which is the comparison yesterday’s page sent you here to make.
The other number is zero. It was zero then and it’s zero now. Nothing sits in Other that you’d have answered today. That’s the number that decides, and the reason is arithmetic rather than principle. Half an hour a day is a bad trade if it costs you one client’s message every two weeks going somewhere you never looked. You’d hand it back inside a month. The zero is the thing that makes the half hour yours to keep.
So this morning is the arithmetic and then one decision. What the half hour is made of. What eight zeros in a row are honestly worth. And the single setting you’ve finally earned the right to move.
The half hour and the part you never timed
Take the half hour apart before you spend it. The pieces come out different sizes than you’d guess.
The check behind the sorting is the smallest piece. It fell from a minute to forty seconds on the third morning and has stayed there. Forty seconds is a small part of a half hour.
The pass down the feed is a little more than that. Most of what it gained came from you stopping bracing before each line, rather than from anything changing on the screen.
Nearly all of the half hour is the opening. On the first sorted morning you read the line Point had written and agreed with it. Then you opened the thread underneath to confirm that you agreed with it, and closed it again, thirty times before nine o’clock. That was the correct behavior for day two, and it was the whole of the time you didn’t save that morning. It’s gone now on everything except the handful you open deliberately. It went because the summaries turned out to be worth acting on, which you established by acting on them rather than by being persuaded.
Now the part worth being straight about, instead of claiming the round number.
The morning you should really be measuring against is the Wednesday before you connected, and nobody timed it. You started writing numbers down on the first sorted morning, which was already on the far side of the largest single change in the two weeks. Whatever the sorting itself saved you sits outside your own measurement, in an ordinary week you kept no record of. The temptation is to fill that in from memory, and memory will hand you forty minutes, because forty minutes is the figure people say about their inbox. That one is a guess. Keep it in a separate column from the half hour, which is evidence.
So the honest account is a measured half hour a day, and an unmeasured amount on top of it. Keeping those apart is worth the small trouble it costs. In six months somebody will ask you whether this was worth doing, and the answer you want to be able to give is the small one you can stand behind.
One test before you believe your own number. What did you do at 8:50 this morning? Day two predicted the two ways people quietly spend the time back, which are reading the inbox anyway and tidying something. If you opened Gmail and worked down it for twenty minutes, the half hour changed clothes, and the paper is measuring the wrong thing. If you did the piece of work you decided on last night, the number is real.
What eight zeros in a row are worth
Now the number that decides. It needs handling carefully, because it gets over-read in both directions.
What you have is eight mornings on which you ran a complete check of everything Point set aside, and found nothing in it you’d have answered. Complete rather than sampled. Every message Point decided against was in one view, and you looked at all of them. That’s a better account of your own sorting than you’ve ever had, including the years when you were doing it yourself and the evidence was a feeling on a Friday.
Proof that Point misses nothing is a different thing, and eight mornings can’t give it to you. Eighty can’t either. A run of zeros is a statement about two weeks, and two things thin these particular two weeks further.
The first is that you were watching. You checked every morning, on purpose, on days when you were interested in the answer. That’s precisely the condition that will slip in November.
The second is that the hard case may never have turned up. Day two named the message most likely to go astray. It’s the first approach from somebody who has never written to you before, where there’s no history for anything to lean on. Go back over the two weeks and ask whether one actually arrived. If no stranger wrote to you in eight working days, your zero is thinner than it looks, because it’s still untested on the case it was always most likely to fail. What a reader of the message alone can and cannot judge about a sender it has never met is AI email triage, and whether you can trust it.
Which leaves you needing a better reason to be comfortable than the one you have, because “it hasn’t missed anything yet” runs out in the winter.
Here’s the reason that lasts. You’re running a setup in which a miss is visible within a day and costs forty seconds to look for. That’s a much smaller claim than a system that never misses, and a far more durable one, because it holds whether or not a good run continues. It depends on a check staying cheap, and the check has cost forty seconds for six mornings running.
That turns the safety question into a question about your habits rather than about the software, and it has a plain answer. Keep the check, daily, while it still costs less than a minute. The morning it starts to feel like a formality, keep it anyway. Put a date on the calendar beside the two you already set last week, and run it deliberately from then on. Write the count down each time, in one column, so that a change is something you can see rather than something you have to remember noticing.
The setting your draft tally points at
Since Friday of last week you’ve been marking every drafted reply one of three ways. Sent with nothing changed, sent after a line or two came out, or abandoned and written from scratch. Read the tally now, and read it properly, because the finding sits underneath the headline proportion.
Say it came out the ordinary way. Most of them needed a line changed. A handful went out as they stood. One or two were rewritten from nothing. Now do the thing Friday said the tally would really turn out to be about, and look at which replies sit in which pile. The untouched ones are the confirmations and the acknowledgments and the yes-that-works-see-you-Thursday. The rewritten ones are the fee conversation and the one where somebody was unhappy with you. Almost every tally comes out that way.
That matters because of the shape of the setting. The autonomy dial is held separately for each kind of work: filing, drafting, scheduling and chasing. It is not held separately for each correspondent. So moving drafting to the top of its range means drafting everything, the fee conversation included. Read that way, a tally that looks like an argument for handing drafting over is a record of the opposite. Your edits are concentrated in exactly the mail you can’t afford to get wrong. Setting how much your inbox does on its own sets out the three positions and what each one lets through, and the week that decision does become live is you decide what gets sent, and when that changes.
So drafting stays where it’s been since the afternoon you connected. Which leaves the question of whether anything moves today. There’s a good answer, and it’s quieter than the one you were looking forward to.
Move filing.
It’s the least exciting of the four and it’s the right one, for two reasons. You have more evidence about it than about anything else. You’ve been filing from summary cards since the third morning, dozens of times, and you can probably count on one hand the threads you’ve gone back for. And its failure is the cheapest in the building. A filed thread has moved inside a mailbox you hold. It’s archived rather than deleted, sitting in Gmail exactly as it sits in Point. One click brings it back, and undoing what Point did is precise about that. Compare that with the only thing in these two weeks that left the building. That was Friday’s reply, and no arrangement of settings could have recalled it.
At the top of its range, filing happens on its own. What you get in place of the asking is the record. It holds what was filed, when, and on what reasoning. It’s the log you went and found on the second morning, while it still had almost nothing in it. It has eight days of your own two weeks in it now, and it’s worth opening once today for that reason alone. The activity log behind every action is what collects there.
Scheduling is the other real candidate, since Point runs the back-and-forth itself. The reason to leave it this morning is that two weeks of ordinary mail may simply have held too few meetings to judge it on. If that’s your week, wait for a busier one. And if the honest answer is that nothing has earned a move yet, leave the dial where it is. That’s a result too. The first kind of work you hand over, and the evidence behind it, is treated as the decision it is in handing over one kind of work at a time.
Whatever you move, move one, and write the date beside it on the paper. The method that got you a legible two weeks stays right on the day the two weeks end.
The rules you were told to leave alone
Monday of last week asked you to leave your filters, your folders and your flags alone. Evidence was coming, and changing one of them mid-way changes the exercise entirely. The two weeks are up and the evidence is in. There are three questions, and all three are answerable now.
Which of your rules has the ranking made redundant? A rule that files a supplier’s mail into a folder you haven’t opened since Monday is kept alive by one thing, which is the fact that you wrote it in 2019.
Which of them hid something? Day two named this trap, and it’s worth checking rather than assuming. A filter that archives on arrival takes a message out before anything else sees it. If a thread went astray in these two weeks and turned up in a folder, you’ve found one. The useful thing to notice is that it was costing you long before Point arrived.
And yesterday’s, which is the real one. Are you now keeping two lists? You were asked to watch which of the two you reached for when something needed writing down. Whatever you reached for is the answer, and it’s worth more than any argument either way.
The instruction is to be selective rather than wholesale. Retire the ones you can give a reason for out loud, leave the rest, and look again at the end of the quarter. Whatever you retire, the labels, the archive, the mailbox and the rules are all still your provider’s. how to switch to an AI email client is straight about which parts survive if you ever go the other way, and which do not.
The hold on your phone expires today as well. That was always about keeping the two weeks to one new variable a day, and there are no variables left to protect.
The half of the paper that is not about Point
Turn the paper over and read the entries that were always about something other than the software.
Yesterday’s split: four dates somebody actually gave you, five worked out from the way a sentence was written, two with no date in the message at all. Underneath it, the tally of obligations you were carrying that had reached you by some route other than mail. Above both of them, Thursday’s dozen names, half of them chosen precisely because they write to you rarely.
Read those three together and they say one thing, and it has nothing to do with an inbox. Your business runs on things said once, in passing, by people who assumed you would remember, and for eleven years the entire record of it has been in one head. You’ve been running it the way every small firm works. That’s why the good ones are exhausting to run.
What’s useful about seeing it written down is that it tells you where the risk has moved to. It has moved out of your inbox, or at least it has changed shape there. The mail arrives, gets ranked, gets read, and gets lifted into a list with a date on it. You have eight days of checking behind that sentence. The exposure now sits almost entirely in the channels nobody is reading. The price rise mentioned on the phone. The favor agreed in a doorway. The thing you promised in a parking lot on the way to somewhere else.
Two habits follow from that, and they’re the only new ones this page asks for. Type those in as they happen, in the seconds after they happen. That’s deliberately about as expensive as thinking of them, and it’s the subject of adding a task in five seconds. And when something that matters gets agreed by telephone, send the one line confirmation afterwards. That used to be a courtesy and good manners. It’s now the act that puts the thing into the channel that’s being watched, which makes it worth doing on a Tuesday when you’re busy and can’t be bothered.
The two things two weeks cannot settle
Two entries on the paper are still open, and both were always going to end the two weeks that way.
The watch has stayed quiet. Eight working days, one watch, no bell. Underneath it the checking marks run five, then four, then three this morning, which is a decline, and a decline is all it can be yet. Day six was clear that the checking stops on the evidence of one fire arriving at the moment you needed it. No fire has arrived, so the reflex has had nothing to be argued out of.
And the names. Half of Thursday’s list was chosen for writing rarely, and rarely has held. Several of them have sent you nothing at all, which is exactly what you were told to expect, and is still slightly unsatisfying to look at.
Both of those are the design working, and it’s worth saying plainly why they sit together on the same page. The two changes you can’t yet judge are the two built for the rare and expensive event, and two weeks of ordinary Wednesdays is the wrong length of time to judge either of them in. You’ll find out about both on a day nobody put on a calendar.
So leave both alone. In particular, leave the watches at one, rather than setting four more so that something happens. Manufacturing evidence by making the software noisier is the same error as reading the inbox to feel busy, wearing better clothes. And leave the quiet names on the list. Quiet was the qualification.
There is no day nine
Tomorrow is Thursday and it’s empty. No screen to find before you let yourself read the feed, no list to write out longhand before you open anybody else’s, no single change to make and then leave alone until Friday. The method is over, and it was built to be over. One change a day with the date beside it is how you make two weeks legible. It’s a method for two weeks, and ordinary weeks run looser than that.
Three things are worth carrying out of it, and they’re all small. The forty second check, daily until it’s dull and then on a date you’ve chosen rather than on a mood. One change at a time whenever you change something, so that the next surprise still has a single parent you can name. And the two numbers again at the end of the quarter, taken on an ordinary Wednesday, against the same clock.
There’s a good deal of Point you’ve yet to meet. Eight days of measurement covered the parts you could put a number against, and skipped almost everything that only turns up when the mail requires it. From tomorrow you meet those the ordinary way, which is the better way round. everything Point does is the inventory, for the week when you’re curious rather than checking.
The paper goes in the drawer rather than the bin, and sentiment has nothing to do with it. It’s the only record anywhere of what your working week looked like on either side of this. In six months, when somebody asks you whether it was worth it, you’ll want the handwriting rather than the impression.
Then, some afternoon in the fall, on a day with nothing in it, the bell picks up a badge and it’s the Hendersons’ letter, weeks after you stopped watching for it. You read it, and answer it, and go back to what you were doing. Nothing about it feels like anything at all, and that’s the whole of what the two weeks were for.