How to write down decisions made in chat

A decision made in chat survives only if someone writes it down as its own short note, in a place that is not the conversation, on the day it happens. Chat has no ending. Nobody stands up, nobody owes the room a summary, so the moment passes with no receipt.

The decision stays where it was made, buried in a run of messages that also holds two jokes and an idea somebody abandoned. It stays findable for about a week.

Most teams do not have a documentation problem. They have a noticing problem, and the fix takes under a minute if you catch it while the thread is still warm.

Why chat eats decisions

Chat eats decisions because nothing marks the moment one gets made.

A meeting has edges. It ends, and the ending is a prompt to write the recap. A channel just keeps going, and the thing you settled slides up the screen at the same speed as a lunch order.

Size is the second problem. The decision itself is often four words long and looks like ordinary agreement. "yeah, let's do that" can commit a team to six weeks of work, and it does not look like anything at all as it goes past.

Then there is search, which everyone treats as the safety net. Search the channel for the topic and you get the whole argument back: every wrong turn, every half proposal, and somewhere in the middle the answer, looking exactly like everything around it.

Search is not a record. A record is something you read in ten seconds and act on. Searching a channel is digging, and digging only works when you already know roughly what the answer was, which means it fails exactly when you need it.

The tell that a decision just happened

A decision just happened when the conversation stops arguing and starts planning.

The signs are boring and reliable:

• Someone writes "ok, let's do that" and nobody adds anything after it.
• The two people who disagreed go quiet at the same time.
• The next message is about who does it, or when it starts.
• You feel a small drop in tension, which is the most accurate of the four.

That is your cue and it lasts a minute or two. Write it now, because tomorrow you will remember the outcome and lose the argument that produced it.

If the thread trails off with no drop in tension, nothing was decided, whatever it felt like at the time. That is a different problem, and asking for the decision plainly fixes it faster than another round of opinions.

Put it where you would look for it

Write the note in the place you would go looking six months from now, which is almost never the channel where it was decided.

Pinned messages feel like the answer. They are not. Pins pile up, nobody prunes them, and after a year the pinned list is a museum of things that were urgent once.

One boring file is enough. A doc or a text file in the repo, named something so obvious a new person finds it without asking. Newest entry at the top.

Do not go shopping for a tool. Once the record needs categories and a template and an owner who maintains it, adding an entry stops being a twenty second job, and anything that stops being a twenty second job stops happening.

Then drop one line back in the channel saying where it went. A record nobody knows about is a diary.

Write the reason, and what would change it

A decision recorded without its reason gets reopened from scratch every time somebody new looks at it. Write the reason down and they can check it instead, which takes a minute rather than a meeting.

The shape that works for me is four sentences: what we are doing, why, what we turned down, and what would make this worth reopening.

Almost nobody writes the fourth sentence, and it is the one that ages best. Reasons expire at a different speed than decisions do, and naming the condition lets the next person test it in a second instead of restarting the whole debate from zero.

A real example from Unvent: three tone options rather than five, because a fourth choice makes picking slower than just fixing the sentence by hand. That reason still holds. If it ever stopped holding, the line tells me precisely what to look at, and the whole thing has never needed a meeting.

It also gives you a way to sort load-bearing decisions from stale ones. When the reason reads as obviously dead, the decision is open again, and everyone can see that without asking permission.

Who should write it

The person who most wants the decision to hold should write it down.

Not the newest person by default. Not a rotating scribe. Scribes produce notes nobody reads, because the writer has no stake in whether the thing survives contact with next month.

If you lost the argument, writing it up is still in your interest. The note carries the reason, and the reason is the part that gets tested later, so a decision you disagreed with, recorded honestly with its condition attached, comes with a built-in review date that nobody can accuse you of scheduling.

There is a failure mode here where writing the note becomes a small power, and whoever holds the pen shades what was agreed. Keep the note dull enough that shading has nothing to grip. Flat statements with no adjectives, posted where the people who were in the thread can push back. Being corrected in writing is cheap while the thread is warm and expensive once the file has sat unread for a quarter.

The decisions you never wrote down

Write the old ones down today, dated today, and say in the note that you are reconstructing from memory.

Never backdate. A date that says when you actually wrote something is worth more than a date pretending the record was there all along.

"As far as I remember, we decided X because Y, and nobody has argued with it since" is a genuinely useful sentence. It is honest about its own reliability, and it hands anyone with a better memory an obvious place to correct you.

Start with whatever is being re-argued right now. Those have a live bill attached. The rest can stay lost, and I mean that: a decision nobody has questioned in a year is not costing you anything, and hunting it down is housekeeping dressed as work.

The strange part is what the habit does to the arguing. Once people know the reason will be written in one plain line, arguments get shorter, because a reason that cannot survive a single sentence was never a reason. It was a preference wearing a coat.