By the second week of December the mail goes quiet. Corinne has never worked out what to do with it.
Her practice is seven people. Behind them sit about four hundred individual returns and sixty business ones. From the middle of February to the middle of April her own mailbox takes three hundred messages a day. In the flat months it takes a hundred and ten. December drops under that. The extensions were done in October, most of the year-end work is sitting with the bookkeepers, and for roughly three weeks nobody outside the firm is waiting on her inside the week.
She’s always used those weeks to rest, which is fair, and to tidy, which feels like getting ready. Then the second week of January arrives, and every arrangement the firm runs for the next four months turns out to be whatever arrangement it happened to have on the last working day of December.
That’s the subject here, and it belongs to December rather than March. It’s about what one practice can settle in the only weeks of the year when settling something is free. And about which of those things are worth the December they cost.
Whether to let software read client mail at all is a separate decision, and an earlier one. Corinne made it in the fall, on the six questions in why you’re right to be careful with AI. This series starts on the other side of that. What’s left is a question of timing.
The four weeks with slack in them
December is full, and it’s worth being accurate about that before treating it as spare capacity.
The clients with a December year end are closing their books. That’s its own body of work, with its own inputs, and March can’t recover them. The job lives in year-end close. The fourth estimate falls due in the middle of January, so the reminders go out now and the silence that follows them starts now. That’s taken apart properly in quarterly estimate reminders. Engagement letters are going out or about to. All of it is real work.
What December has that no other month has is slack of a specific kind. The hours are full. What’s different is that nobody outside the firm is waiting on you within the week. Every December job has a date, and every one of those dates is further out than tomorrow. So a morning spent on work that pays off later is free today.
That matters, because the month you make a change in decides how the change goes. A change to how the mailbox works, made in the first half of December, has six weeks. Six weeks to settle, to be wrong twice, and to be corrected before anything depends on it. The identical change made on the tenth of February is a second emergency laid on top of the first. It gets abandoned by the fourteenth. The firm concludes the thing doesn’t work, when what happened is that it started in the wrong month.
Firms know this. The difficulty is that nothing in December forces anything. Every other decision a practice makes gets made because a date arrives and makes it. December has no such date. So the month goes on the small satisfying work of clearing a screen. The screen clears. January arrives the same as it would have.
So the December discipline is a strange one. You treat the missing deadline as the reason to act.
What last season left in your sent items
Start with what went wrong last year. The record is already there, in your own mail, and it’s better than what you remember.
Memory of a season leans in a particular direction. You keep the one disaster, told as a story, usually with a villain in it. The eleven small ones fade. Each cost twenty minutes and a little credibility, and each was too dull to survive April.
The record sits in your mail, written in your own hand. Corinne ran four searches on her sent items, last January through last April. She looked for the phrases she uses when she’s been slow: apologies for the delay, only just seeing this, resending, thanks for your patience. Thirty-one messages came back. Then she searched what she’d received, for the phrases clients use when they’re chasing: following up, checking in, second time, did you get my.
The second list is the better one. Thirty-one apologies tell you what you noticed. A chase tells you what mattered to somebody paying you. That list is shorter and more expensive, because people only chase twice for things they care about.
Then look at the shape rather than the count. Corinne’s thirty-one clustered inside those four months. Nineteen fell inside three weeks in the middle of March, which tells her about capacity. Seven of the rest went to the same two clients, in weeks that were quiet, which tells her something else. Those two send everything as a forwarded chain from their bookkeeper, with nothing written above the attachments. A message with no words in it looks like a message that needs nothing.
Those two findings want opposite responses, and telling them apart is most of the value of the exercise. A cluster in mid-March is volume, and attention alone leaves volume where it is. A pattern attached to one sender, or one kind of message, is a hole in how the mailbox surfaces things. It’ll be there again in June, with no season to blame. The general case is that a flat list sorts on how loudly a message was written, rather than on what it will cost you, and email overload works it through. It bites hardest in a practice because a client’s mail is only readable against the list of what that client still owes you, which running the firm inbox sets out.
An hour on this, once, in December. It’s the only hour of the year when the evidence is complete and nothing’s on fire.
The list of names is a year out of date
Every practice has a list of people it can’t afford to miss. In most practices it lives in one person’s head, and it was last revised by accident.
A year has passed since it was accurate. Clients have left. Clients have arrived. The ones who arrived in November have no history in the mailbox at all, and that’s exactly the condition that makes a message look unimportant. A long-standing client changed bookkeepers in August, and the new one has written to you twice. The bank moved your relationship manager. A client hired a controller, and from January that controller is the person who answers, in place of the owner you’ve dealt with for nine years.
Then there’s a category that only exists between January and April, and it’s the one nobody writes down. A payroll bureau that sends the year-end reports from an address no human reads. A brokerage’s document team. The portal notification from a state revenue department. It arrives once, it looks like every other automated notice, and it’s the only warning you get. These are bureaus and agencies rather than clients. Several of them decide whether a return can be finished. Their history in your mailbox gives you no reason to think they matter.
The useful move is to build the list from the engagement list rather than from recollection. Whoever you’ve accepted for the season is on it. Then, for each of them, whoever else writes on their behalf. That second half is where the misses live. A firm’s instincts are trained on client names, and half the mail that decides a return comes from somebody else.
What naming people in advance does, and the honest limits of it, is covered in VIP mail that never gets buried. The sitting-down-and-doing-it version is day four of the first two weeks, including the ten minutes on the contact list that turns the exercise into something a machine can use. December’s part is smaller. The list is stale, the staleness is invisible, and January is when it costs.
The mail you are about to cause
Here’s the part practices consistently miss. It’s the most predictable inbox event of the entire year, because you schedule it yourself.
Somewhere between the middle of December and the middle of January the firm sends its season mail out to the whole base at once. The organizer. The engagement letter. The fee change, if there is one. The portal invitation. For Corinne that’s four hundred and sixty near-identical messages inside two weeks. It’s the single largest thing her firm does to its own mailbox all year.
What comes back arrives in three shapes. Each one is a different job, and the file you asked for is rarely any of them.
Some of it is questions. About the fee. About whether they still need to fill any of this in, given last year. About something on the prior return they’ve been meaning to ask since June. Every one of those is unscheduled work with no date attached, which is the whole reason a question pile behaves the way it does. That lives in client questions.
Some of it is an early delivery. One form, sometimes one photograph of one form. It arrives months before the rest of that client’s file, and it matches nothing, because the list it belongs against doesn’t exist yet.
And some of it is silence. That’s the largest of the three. It has to end up somewhere with a date on it, or it turns into a client you’re expecting who never arrives.
The arithmetic is worth doing before the send rather than after. Four hundred and sixty messages, at a fifteen percent reply rate in the first ten days, is roughly seventy replies. Most need a human decision. They land in the same week, on top of the fourth-estimate run and the December closes. That’s a March-shaped week, in January, and it happened because everything went out on a Tuesday.
Corinne’s decision this year was to stagger it. Businesses first, in the second week of December, because their deadlines come in March and their replies are the ones that unblock the most work. Individuals in three alphabetical batches across January. The clients themselves neither notice nor care. She does it because the replies arrive in proportion to the sends. Her capacity to answer them is flat. The sending is the only variable she controls.
The signature half of that batch is quieter and slower, and it blocks nothing until it blocks everything. It belongs to engagement letters. The requests half belongs to chasing client documents, along with the six or so reasons a document runs late, of which forgetting is only one. December decides when they land. How to do either is a question for another month.
The change that cannot be made in March
If the firm is going to change how the mailbox works, this is the window, and the reason goes beyond having the hours.
Anything that ranks your mail learns your practice from your mail, over weeks. Which senders you answer inside the hour. Which threads you leave for Thursday, and are right to. What you mean when you write back in three lines. That learning happens somewhere. In December it happens across two weeks where a wrong guess costs nothing, because the messages it gets wrong are messages with no date on them. In February it happens while you run the season, on something you’re still learning to read.
Corinne connected Point in the first week of December. She picked the first week because the last one is the holidays and half the firm is out, and two weeks of learning needs two weeks of ordinary traffic to learn from. Point runs on top of the mailbox the firm already keeps, so there was no migration, and no new address to tell four hundred clients about in the worst possible month to tell them anything. The two weeks that follow are walked day by day in connecting the inbox you already have.
What she wanted in place by the second week of January was narrow. Mail ranked before she sits down, so the first screen of a season morning is a judgment rather than a delivery order. A plain summary on a thread before she opens it, which matters most on the forwarded chains from bookkeepers that carry no words of their own. An ask inside a message becoming a task with a date on it, so a promise made on the fourth of March sits somewhere other than the head already holding forty others. Point’s full inventory sits on benefits. Practices get their own version of it in AI email for accountants.
She set one thing deliberately low: how far any of it goes on its own. The dial is per kind of work rather than one setting for everything. Out of the box it sits at review, which is the right place to run a first season from. Filing and ranking can be left to get on with it. Anything that writes to a client is worth watching through a first March before you move the dial.
Two honest notes belong here, because December is when they’re cheap to act on.
The first: a light month teaches a light-month version of your practice. What Point knows on the second of January is Corinne’s December, and her December is a different month from her March. The corrections that matter most come in the first two weeks of the season, when the mail changes shape and the ranking is still catching up. Connecting early leaves that intact. What it changes is that the correcting happens on a system you already know how to correct.
The second is a boundary worth naming plainly. Some firms need a status list per client, showing which of eleven documents have arrived and been accepted. An email client isn’t that, and no amount of ranking makes it that. Email carries requests well and holds registers badly, which is the whole argument of one PBC list rather than five copies of it. A practice can want both things, and it’s worth knowing which one you’re buying. Point changes what reaches you and in what order. That’s a real problem in a season. Knowing where every file stands is a different problem.
What the second week of January does not care about
Corinne got through most of it. The sent-items hour, done on a Thursday afternoon. The list of names, rebuilt from the engagement list: ninety-one entries where last year’s had sixty-two, thirty of them people who have never once emailed her. The organizer batch staggered. Point connected on the fourth, wrong about her for eight days, sensible about her by the fifteenth.
One thing she left unsettled, and she knew it at the time. Four people in the firm now write to clients. A client answers the person whose name was on the last message, which is sensible behavior, and nobody’s going to talk them out of it. So where does the record of a request live when the request went out from somebody else? She left that question in December. She told herself she’d take it in the first quiet week of the season. The season has no quiet week.
That’s the shape of it. December runs out rather than finishing. Whatever got decided is what the firm has. Whatever got left over meets the firm again in March, at four times the price.
The calendar consults nobody about any of this. The first forms of the season arrive well before any deadline requires them. They come from the clients who are early rather than the ones you’re thinking about, in ones and twos, often forwarded with nothing written above them.
The first one to land on Corinne’s screen arrives without saying which of the two jobs it belongs to. There’s the pile the practice has to collect. There’s the pile the practice has to issue. They arrive in the same two weeks, under the same word, from senders that look alike, into one mailbox. Sorting one from the other is the first real work of January. Getting it wrong is expensive, and the cost shows up at the end of the month.