The Slack-Thread Graveyard: Where Answers Go to Be Asked Again

Somewhere in your chat history is a beautiful answer. Someone asked how to fix the export bug, or what the discount policy is for annual plans, or which vendor handles the badge printers — and someone who knew took ten careful minutes to write it all out: the steps, the exceptions, the login nobody remembers. It was accurate, complete, and generous. It survived about a week. Then it sank below the scroll line, where chat search — which needs the exact words the original asker happened to use — will never quite surface it again. Eight months later the same question arrives fresh, and the same expert types it all out again, slightly worse, from memory.

That's the graveyard: not a place where knowledge is missing, but a place where it was captured in a medium designed to forget. Chat is a conversation surface, and conversations are supposed to scroll away — that's what makes chat fast. The failure isn't using chat to answer questions; everyone will always answer where the question was asked. The failure is having no ritual that moves the answer somewhere permanent before the scroll takes it.

The answered-twice rule

You cannot document every thread, and shouldn't try — most chat answers are one-time coordination, dead on arrival by design. The filter that works is memory-light and judgment-free: the second time you answer the same question, it becomes an article. One occurrence is trivia. Two is a pattern, and a pattern has a price: every future recurrence costs an expert's interruption, a requester's wait, and — the expensive part — the drift that creeps in as the answer gets retyped from memory. The second ask is the signal that this knowledge has a career ahead of it. Catch it there.

Ten minutes, four parts, written by the answerer

The capture has to be cheap or it won't happen, so the template is four parts: the question as people actually ask it (the title, in the asker's words — not your taxonomy's words — because that's what the next person will search); the answer, steps first; the exceptions and gotchas that made the chat answer long in the first place; and the owner and date, so the next reader knows how much to trust it and who to ask when it ages. Rough is fine. A scrappy article that exists beats a polished one queued behind good intentions.

And it's written by the person who just answered — not routed to whoever owns documentation. They have the answer paged into their head at that exact moment; anyone else starts by re-deriving it, which quadruples the cost and usually kills the capture entirely. Ten minutes from the person who knows, immediately, while the thread is still warm.

Link, don't retype

The habit that makes the system self-reinforcing: from the day the article exists, the answer to that question in chat is a link. Question appears, article link lands, plus one line of human warmth so it doesn't feel like a ticket deflection. This does three jobs at once: the asker gets a better answer than an off-the-cuff retype; every link teaches the team the knowledge base is where answers live — which is what finally gets people searching it before asking; and if the article is wrong or stale, the link surfaces that immediately, in public, where the fix is a two-minute edit instead of a vacation-week discovery. If you find yourself typing a real answer into chat for the third time when an article exists, that's not a discipline failure — it's the article telling you its title doesn't match how people ask.

The weekly sweep

The rule catches most of it in the moment; the sweep catches the rest. Once a week, ten minutes, someone skims the help channels for answers that deserved capture and didn't get it — the tell is any reply over four sentences long. Rotate the job. The sweep isn't for writing the articles; it's for tagging the threads and nudging the answerer — “that reply about the badge printers is an article, you've written it twice this month.” Ten minutes of curation keeps the rule honest without turning anyone into a librarian.

The bottom line

Your team already writes documentation every day — carefully, accurately, in a medium built to lose it. The fix isn't a documentation initiative; those die of ambition. It's one rule, one template, one habit, one sweep: answered twice means it becomes an article, four parts in ten minutes, written by the answerer while it's warm, and from then on the chat answer is a link. The graveyard keeps filling either way. The only question is whether the knowledge gets out before the scroll closes over it.

— Tom