It is Friday afternoon and an account lead at a twelve-person agency is doing archaeology. The client update is due at five. The design decision is in a WhatsApp thread from Tuesday. The revised deadline is in an email she has to search for twice. The open question about the invoice is in a comment on a shared doc that two people have seen. She assembles all of it into four paragraphs, sends it, and knows she will do the same excavation next Friday.
We think this ritual deserves more suspicion than it gets. Nobody questions that the update has to be written, because someone has always written it. But look at what the work actually is: not judgment, not communication skill — retrieval. Every fact in that email already existed. It was just stored in five places, none of which could answer a question.
The reconstruction tax
For a team of five to fifty people, this tax lands on whoever is closest to the client. An account lead who spends four hours a week assembling updates spends two hundred hours a year on retrieval. That is five working weeks of finding out what her own team already knows.
The standard answer is a project management tool, and the standard result is a sixth place where information lives. Boards and cards are good at showing state — what column a task is in — and bad at holding history: why the deadline moved, who agreed to the scope change, what the client said when they saw the draft. So the chat about the work stays in the messenger, the decisions stay in email, and the board becomes one more thing to reconcile on Friday.
Why the chatbot didn’t save you
The newer answer is an AI assistant bolted onto the same tools, and it disappoints for a reason worth understanding. An AI can only summarize what it can read. A decision buried in a comment thread on a color-coded card, or implied by a card moving between columns, is close to invisible to a language model. The tools were optimized for human scanning, and the price of that optimization is that the machine can’t follow the story either.
This is why we built Otopic text-first. Work is stored as structured, human-readable text: the note is the durable source of truth, decisions are short readable paragraphs recorded where they were made, and chat is attached to the work it is about. A container like that — we call it a Topic — accumulates a history the AI can actually parse, quote, and link.
The report becomes a query
Once the work has memory, the Friday ritual inverts. Instead of a person reconstructing the story, the AI drafts the update from what actually happened in the Topic: what moved, what stalled, what was decided and why. The account lead reads it, adjusts a sentence, and sends it. Better still, the client often doesn’t need the email at all — a client-facing view of the Topic answers “where are we on this?” before anyone asks it.
There is an honest trade-off here. Text-first storage means Otopic will not win a screenshot contest against a beautiful board, and teams who love arranging cards will notice the difference. We accepted that cost deliberately: visual state is a read layer you can always add on top of structured text, but structured text cannot be recovered from a pile of visual state. The decision ledger that makes the AI reliable is the same one that makes your work portable and readable in ten years.
The account lead’s two hundred hours don’t disappear — they go back into the work the client is actually paying for. The status report still exists. It just stopped being anyone’s job.