A question that keeps coming back is usually not a search problem. It is a decision that was made in a conversation and never written down. Search cannot find something that does not exist, so a search layer leaves those questions exactly where they were.
Someone asks in the team chat whether you charge for a same-day callout. Three people answer with some version of “I think so.” Eventually the person who has been there longest gives the real answer, with the reasoning: yes, unless it is a maintenance client, and never when the return trip was our mistake. Everyone thanks her and the thread moves on. Four months later the same question comes back, worded slightly differently, sometimes from someone who was in the first thread.
It is worth paying attention to which questions come back. The ones asked once, answered, and never asked again have often been written down somewhere. Someone found it, or was pointed at it, and the matter closed. The ones that resurface every few months are a different animal. Nobody is failing to search properly. There is nothing to search. The answer was decided in a conversation, applied correctly ever since, and never written anywhere a second person could reach.
Part of the reason is that a decision does not feel like a document. Pricing exceptions, which supplier to use when an order is urgent, whether a long-standing client keeps the old discount: these get settled on a call, by someone with the standing to settle them. That was real work. It just left almost no trace, apart from an email to one customer and a memory in one head.
So the instinct is findability. Better search, a wiki, an assistant pointed at the shared drive. That can genuinely help, and before you buy it is worth checking whether you already pay for something that does most of the job. But retrieval only reaches what somebody already wrote. Pointing a capable tool at your files makes the documented questions faster to answer and leaves the re-asked ones untouched. The repeat rate is the tell. A high repeat rate points at decisions that live with people rather than anything wrong with the people.
The more useful move usually costs little beyond attention. Keep a running list of every question that gets asked more than once, and note who ended up answering it. Each one is a decision the people who need it cannot easily reach. Then capture the reasoning, not just the verdict, because what people need is the rule, the exceptions, and who decides when it is genuinely unclear. Knowing which parts are a rule and which are a judgment also tells you how much of it technology can safely carry, and how much has to stay with a person for now. Next time a familiar question shows up in chat, answer it, then ask one more thing. Has this come up before?
See how we clarify the right path
See how we clarify the right path
About the author
Missy Ross
Missy Ross is the Founder and AI Architect of Vero Dawn, an AI architecture and solutions company that helps businesses see what could work better, determine the right solution, and bring it to life. Before founding Vero Dawn she spent 21 years in internal audit across banking and manufacturing, at Audit Director level.