The Answer to Sales' Question Is in Support's Ticket History

The knowledge a company is missing almost always already exists somewhere inside it. The problem is rarely that no one wrote the answer down — it's that the answer was written down by a different team, in a different system, for a different purpose entirely.

A sales engineer is on a call with a prospect who is close to signing, and the prospect asks the one question that always seems to arrive at the worst moment: does the product handle a particular edge case in their environment, the kind of specific integration wrinkle that no data sheet anticipates. The sales engineer does not know, because it is genuinely obscure, and so they say the thing that everyone on that side of the table says a dozen times a week — let me check with the team and get back to you. They will send a message to a channel, wait, escalate to a product manager, wait some more, and eventually surface an answer three days later, by which point the deal has cooled a degree or two. What none of them realizes is that the answer was fully resolved eight months ago. A support engineer worked the exact same edge case with an existing customer, diagnosed it, documented the fix in careful detail, and closed the ticket. The knowledge exists. It is written down. It is precise, tested, and correct. It simply lives in a system that the sales engineer has never had a reason to open, filed under a customer name they have never heard, in the closed-ticket archive of a team they rarely talk to.

This is the shape of almost every knowledge gap inside a large organization, and it is worth naming clearly because companies keep misdiagnosing it. The information the sales team needs is not missing. It is exhaust — the byproduct that another team generated while doing its own job, sitting exactly where that team left it, invisible to everyone who did not produce it. Sales' answer sits in support's closed tickets. Support's answer sits in engineering's pull-request discussions. The clause that legal already negotiated and approved sits in a SharePoint folder that no one indexed, three reorganizations ago, under a project codename that has since been retired. The knowledge is there. What is missing is any way to reach across the wall and retrieve it.

The reflex is to write more, and it never closes the gap

When an organization notices these gaps, the instinctive response is almost always the same, and it is almost always wrong. Someone declares that the company has a documentation problem and launches an initiative to fix it. There will be a new wiki, or a knowledge base, or a mandate that every team maintain a canonical FAQ, and for a quarter or two the effort produces a burst of freshly written pages that answer the questions people were asking at the moment the initiative began. Then the pages age. The product changes, the edge cases evolve, the person who owned the wiki moves teams, and within a year the carefully authored knowledge base is a museum of answers to last year's questions, consulted by no one and trusted by fewer.

The reason this cycle repeats is that writing more documentation treats the symptom and misreads the disease. The company did not fail to answer the sales engineer's question — support answered it thoroughly, in the natural course of solving a real customer's problem, and the answer was more accurate than anything a documentation initiative would have produced because it came from an actual case rather than a hypothetical one. Asking teams to also write everything down a second time, in a central place, in anticipation of some other team someday needing it, is asking them to duplicate knowledge they have already recorded, for an audience they cannot see, about questions that have not been asked yet. It is enormous, ongoing, thankless work, and even when people do it faithfully the output decays faster than they can maintain it. The documentation project is a tax on the teams that generate knowledge, levied to compensate for the fact that no one can retrieve the knowledge they already generate.

There is a quieter cost hiding underneath all of this, too. Every hour a company spends re-answering a question it has already answered, every deal that slows while someone chases a fact that was sitting in a closed ticket, every contract that takes an extra week because the approved language could not be found and so legal drafts it again from scratch — all of that is the compounding interest on knowledge the organization already paid to create and then could not locate. The company is not short on answers. It is paying, over and over, for the same answers, because the first copy was filed somewhere the second person could not look.

Every team's answer is another team's exhaust

Once you start seeing knowledge gaps as retrieval failures rather than authorship failures, the pattern shows up everywhere, and it is remarkably consistent. The most valuable institutional knowledge in a company is rarely produced by anyone sitting down to produce knowledge. It is produced as a side effect of work — the support engineer solving a ticket, the account manager summarizing a difficult renewal, the developer explaining in a code review why a particular approach was rejected, the finance analyst annotating why an unusual expense was approved. None of these people believe they are writing documentation. They are leaving a trail while doing their jobs, and that trail is the richest, most current, most trustworthy record the company has of how things actually work. It is also, by its nature, scattered across a dozen systems that were each designed to serve a single team, with no expectation that anyone outside that team would ever need to see inside.

The walls between these systems are not primarily technical, though they present as technical. A support platform, a CRM, a document store, a code host, a wiki, a shared drive — each is a well-lit room with the door facing inward. People who live in one room see everything in it and nothing next door, and over time they stop imagining that the answer they need might be sitting one room over, because they have no window into that room and no habit of looking. The organizational chart hardens into a knowledge topology, where information can flow freely inside a team and barely at all between them. The company behaves, functionally, as a set of separate businesses that happen to share a logo, each hoarding — without meaning to — the exact answers the others are burning hours to reproduce.

What makes this tragic rather than merely inefficient is that the raw material for solving it has been present the whole time. The answer to the sales question genuinely is in the support ticket. The legal precedent genuinely is in the unindexed folder. The organization is not knowledge-poor; it is knowledge-rich and retrieval-poor, sitting on a vast, current, hard-won store of exactly the answers it keeps claiming it lacks, unable to reach most of it because reaching it would mean crossing a wall that no tool was ever asked to cross.

Retrieval across the walls is the actual unlock

The capability that changes this is not a better place to put documents. It is the ability to search across every system where knowledge already lives, treat them as one connected body of institutional memory, and reason over what comes back — to take a question posed in sales' language, recognize that its answer is phrased in support's language inside a closed ticket, retrieve it, and understand that the two are the same thing. That is a fundamentally different act from storing a document, and it is the reason the documentation reflex fails: the goal was never to create one more room, but to knock windows between the rooms that already exist. This is precisely what a protocol like Model Context Protocol is for — a common way for an intelligent system to reach into the support platform, the CRM, the code host, and the forgotten SharePoint alike, so that a company's Enterprise Knowledge is defined not by what someone remembered to centralize but by everything the company has ever actually written down, wherever it was written.

It is also the part of the current AI wave that is easiest to get wrong, which is worth saying plainly, because the market is crowded with tools that promise cross-silo intelligence and deliver a chatbot pointed at a single wiki. Gartner has predicted that over forty percent of agentic AI projects will be canceled by the end of 2027, citing unclear value and what it calls "agent washing," and a great deal of what gets sold as enterprise knowledge retrieval is exactly that — an old search box relabeled, still confined to whichever room it was installed in. Retrieval that only reaches one system reproduces the original problem with a friendlier interface. The unlock is not a smarter index of the documentation a company was already failing to maintain. It is a reasoning layer with the reach to gather an answer from wherever it happens to be filed, and the judgment to recognize it as the answer once it arrives. This is the premise beneath the broader move toward an autonomous enterprise: a company whose Autonomous AI Workers and Specialist Agents can draw on the whole of its institutional memory rather than the narrow slice their own team happened to author, so that the sales engineer's question is answered from the support ticket, in the moment, without a three-day relay through people who also do not know.

The reframe worth carrying out of this is that a company's knowledge is not measured by what it has written down, because it has already written down far more than it can use. It is measured by how far any one person, or any one system acting on their behalf, can reach across the walls to retrieve it. Seen that way, the organizations that pull ahead will not be the ones that write the most documentation. They will be the ones that finally stop treating knowledge as something to be centralized and start treating it as something to be retrieved — reaching into the exhaust of every team and discovering, again and again, that the answer they were about to spend three days and a fresh document producing was sitting one room over the entire time, already written, waiting to be found.