Direct messages and groups: keeping internal communication readable
A company of ten people does not have a communication problem in the way a company of a thousand does. Everybody can reach everybody in one hop. Nothing is far away.
What small companies lose is not the message. It is the thing inside the message: a decision somebody made, a date somebody agreed to, a small promise made in passing on a Wednesday. Those are the parts that other people plan around, and they are exactly the parts that a scrolling conversation is worst at holding.
Three ways it stops being readable
The first way is one group for everything. It starts as the company group because there are six of you and a second group would be silly. Then hiring, a customer problem, a bug, lunch and a pricing decision all land in the same hour, and reading it back means reading all of it. Nobody reads it back. The pricing decision is now findable only by the three people who were online when it happened.
The second way is a decision buried in a long thread. Somebody asks a good question, four people answer, positions move, and on the nineteenth message somebody writes “ok let’s do the second one”. That sentence is the only part that matters, it is indistinguishable from the eighteen around it, and in two weeks nobody will be able to say when the decision was made or what exactly was decided.
The third way is the promise nobody wrote down. Two people talk directly, one of them says they will have it ready by Thursday, both of them mean it. It exists in two heads and nowhere else. On Friday the second person is annoyed and the first person is surprised, and neither of them did anything wrong.
All three have the same shape. Something durable was said inside something disposable.
What a direct message is for
A direct message is for the thing that concerns exactly two people. That is a narrower category than most companies use it for, and keeping it narrow is what makes it useful.
It is right for a question with one obvious owner, for anything about a person, and for the early, half-formed version of an idea you are not ready to put in front of everyone. It is right for the sentence that would derail a group thread if you said it there.
It is wrong for a decision. The moment two people in a direct message agree on something that changes what a third person should do, the conversation has outgrown the channel it is in. That does not mean paste the whole thread into a group. It means one sentence, in the place the affected people read, saying what was decided.
The test is simple and worth applying honestly: if somebody who was not in this conversation would have to be told the outcome, the outcome does not belong only here.
What a group is for
A group is for a subject with more than two people in it and an end.
The “with an end” part is the one small companies skip. A group for a customer migration, a group for a hiring round, a group for the release: these have a reason to exist, and a point at which they should stop. A group for “general” has neither, which is why it absorbs everything and stays unreadable.
Two or three subject groups plus whatever direct messages people need is usually the right shape for a ten-person company. More than that and people start posting in the wrong one, which is worse than having fewer. Fewer than that and you are back to one group for everything.
One more thing helps more than it should: say what a group is for in the first message in it. Half of what makes a group unreadable is that its scope was never stated, so everyone made a reasonable guess and the guesses differed.
Turn what was said into something with an owner and a date
This is the habit that does most of the work, and it is not a tool question.
When a message contains a commitment, it becomes a todo, with the person who owns it and the day it is due. Not a summary. Not a note. An item that exists on its own, outside the conversation, that somebody can look at next week without reading the thread it came from.
Do it at the moment it happens. A promise captured in the ten seconds after it is made costs nothing. The same promise captured on Friday requires somebody to re-read Wednesday, and it will not happen.
The same goes for decisions. A decision that stays inside a thread is not a decision, it is a message that several people happen to remember. The version that survives is one line: what was decided, when, and who it changes things for. Where that line lives matters less than that it exists outside the conversation.
This is the whole discipline: conversation is where things get worked out, and it is a bad place to keep them afterwards.
What a chief of staff does with your messages, and where it stops
Software can carry the tedious half of this, and it should be clear about which half.
A chief of staff can read along. After a conversation it proposes what is worth keeping, and nothing is used until you confirm it, which is the shape this habit wants: something you accept or drop in a second. Once a todo exists with a date on it, it follows up. And it can draft the reply, so the message asking whether Thursday is still right costs you one read and one click rather than five minutes of careful wording.
Then it stops. It does not send. Every draft is yours to change or throw away, and everything it writes is signed in purple, so nobody in a group is ever left wondering whether they are reading you or a machine. That boundary is not a limitation to be relaxed later. It is the reason the drafts are useful: a group conversation only stays readable if every message in it is accountable to a person.
Clean the groups once a week
Ten minutes, once a week, keeps this from decaying.
Close the groups whose subject is finished. An archived group is findable; a group with eleven dead subjects in it is not.
Look at anything that got answered in a group and is still somebody’s outstanding commitment, and check that it exists as a todo with a date. Anything that only exists as a message is at risk.
And notice where things were posted in the wrong place. If the customer group keeps collecting hiring chatter, the problem is usually that there is no obvious place for hiring chatter, not that people are careless.
Where Opitor fits
Opitor is an AI-native operating system for small companies, where every person gets their own AI chief of staff. Messages, direct and group, sit in the same product as the Todo List and its deadlines, so a promise made in a conversation becomes an item with an owner and a date without leaving for another tool. Each person’s chief of staff reads along, follows up their todos and deadlines and drafts the reply, and then it stops: it drafts, you send, and everything it writes is signed in purple. Administrators have no override, so a direct message stays between the two people in it. More on that division of labour is on what is an AI chief of staff and the questions and answers, with a comparison to a team messaging tool on Opitor vs Slack.