Why Knowledge Capture Feels Like Homework

Here is what makes this hard to fix. Most people know they should be recording more. They know the call notes matter. They know the CRM should be updated. They know that if they just wrote down what the customer said, next time would be easier.

They also know it feels like homework.

There is something deeply human about this. You finish a call. You have been engaged, present, doing actual work. And then the system asks you to stop doing work and start doing admin. Summarise what happened. Tag it. File it. Link it to the right contact. Update the opportunity stage. Add a note for next time.

It is the professional equivalent of being asked to show your working in a maths exam. You got the answer. You know the answer. But the system does not trust you until you have written it all out in the approved format.

So people skip it. Not maliciously. Just humanly. They move to the next call. The next email. The next thing that actually feels like work. And the information stays in their head.

Every knowledge tool ever built asks busy people to stop doing their job in order to describe their job. That is why the data is always incomplete.

This is not a training problem. It is not a discipline problem. It is a design problem. Every tool built to capture knowledge requires the human to do something. Stop working. Switch context. Open a different app. Type up what just happened in a format the system can digest. And this is being asked of people who are busy, under pressure, and whose priorities do not naturally align with the system's.

Take sales as an example. Turnover among salespeople sits around 35%, nearly three times the rate across all professions. The average rep stays 18 to 20 months. They will not be around to reap the rewards of meticulously recording information. So they optimise for what matters to them right now: hitting quota, closing the next deal, keeping the pipeline moving. The CRM entry that would help their replacement in eighteen months? That is not on their radar. It is not even rational for it to be.

We have built knowledge systems that depend on the least incentivised people in the company to do the most tedious part of their job, consistently, under time pressure, for someone else's benefit. And then we act surprised when the data is incomplete.

The answer has to be passive. The capture has to happen in the background, without anyone needing to do anything. Not "easier logging." Not "fewer fields." Not a slightly nicer interface for the same admin task. The process has to disappear entirely. Knowledge should flow into the system as a byproduct of work, not as a separate chore layered on top of it.

Why CRMs, Wikis, and Note-Takers Have Not Solved This

The enterprise software industry has spent decades building solutions to this problem. None of them have worked. Not because the tools are bad, but because they all make the same fundamental mistake: they put the burden of capture on the human.

CRMs were supposed to be the company's memory of every customer relationship. In practice, they are a compliance system. Reps fill in the fields they are measured on and ignore the rest. The CRM knows the deal stage and the close date. It does not know that the buyer's real concern is internal adoption, or that they mentioned a restructuring in passing that is going to delay everything by two months. That context was on the call. It never made it to the record. Research suggests that up to 80% of opportunity-related data gathered by reps never gets logged.

Knowledge management platforms (internal wikis, company intranets) were supposed to capture institutional wisdom. In practice, they capture what someone had time to write up on a slow Friday. The wiki tells you the process. It does not tell you what happened on the call that changed the process. And the half-life of a wiki page is brutal. Within months it is outdated. Within a year it is actively misleading. Nobody goes back to update it because updating documentation is somehow even less rewarding than writing it in the first place.

Meeting note tools are the most recent attempt. They have got closer to the right idea because they sit inside the conversation rather than after it. But most of them stop at transcription and summary. They tell you what was said. They do not connect it to what it means: which account this relates to, what was promised, what changed, who needs to act on it. A transcript without context is just a searchable record of words. It is better than nothing, but it is not memory. There is a meaningful difference between capturing a recording and building a memory.

The pattern across all of these is the same. Each tool captures a fragment. None of them build the full picture. And every single one depends on a human choosing to use it properly, which brings us back to the homework problem. A company brain cannot be built on tools that are waiting for input from people who are busy doing their actual jobs.

The Knowledge Gap Is a Structural Failure

At the individual level, this is a performance gap. At the company level, it is a structural failure.

The most important context from a sales call, a support conversation, a partner meeting: it lives in someone's head. It stays there until either the person remembers to use it (unlikely, given six months and 400 other conversations in between) or the person leaves the company entirely (at which point it is gone forever).

Now think about what companies actually claim as competitive advantage. It is almost always knowledge. We know our customers better. We understand the market. We have learned from our mistakes. We have institutional wisdom.

But if most of that knowledge was never captured, if it exists only in the heads of people who might leave next quarter, then the competitive advantage is a fiction. The company thinks it knows things. In reality, individual people know things. And the company is just hoping those people stick around long enough.

Companies claim knowledge is their competitive advantage. But most company knowledge was never written down, and what was written down is incomplete. The advantage is built on individual memory, and individual memory does not scale.

Where Customer Relationships Go to Die

This gets worse in customer-facing teams, where handovers are constant and context loss is the default.

Think about how a customer actually experiences a company. A salesperson spends three to six months building a relationship. They learn the buyer's priorities, their internal politics, their concerns about implementation, their communication preferences, the name of the project sponsor's dog. All the soft detail that makes the relationship feel real.

Then the deal closes and the customer gets handed to a customer success manager. In theory, there is a transition. In practice, it is a 30-minute call and a half-populated CRM record. The CS manager gets the contract details, the deal history, the technical requirements. What they do not get is everything that made the relationship work. The buyer's communication style. The concern they raised once on a call in month two. The fact that they hate being surprised in QBRs and prefer a heads-up email the day before.

56% of customers say they have to repeat themselves to different representatives. And this is not a one-time event. Throughout the customer lifecycle, handovers keep happening. Sales to CS. CS to support. Support back to CS. CS to renewals. Each time, more context disappears. Each time, the customer has to re-explain themselves. Each time, the relationship gets thinner.

Nobody can reasonably expect a salesperson who spent six months building a relationship to transfer everything that matters in a one-hour handover call. The soft detail that relationships are actually built on, the trust, the personal context, the unspoken understanding of what this person cares about, none of that fits in a CRM field or a handover template.

And the customer notices. They might not say it directly, but they feel the gap. The new person does not know them. The company that seemed so attentive during the sales process suddenly feels generic. The relationship resets to zero, and the customer starts wondering whether they made the right choice.

What Organisational Memory Actually Means

This is what organisational memory actually means, and it is worth being precise about it because the term gets thrown around loosely.

Organisational memory is not dashboards. It is not a data warehouse. It is not the company wiki that nobody reads. It is the accumulated understanding of why things happened, what was tried, what the customer actually cares about, and what the real tradeoffs were behind a decision.

In most companies, this memory is leaking constantly. Not through some dramatic failure, but through the slow, daily erosion of people not writing things down because it feels like homework. Through handovers that compress months of context into an hour. Through turnover that takes lived experience out the door.

What companies actually need is a shared knowledge layer that sits underneath the work, capturing context as it happens and making it available to whoever needs it, whenever they need it. Not a new tool that people have to remember to use. A memory layer that remembers what the humans cannot.

The distinction matters. A knowledge layer is not a database. It is not a search engine. It is a living representation of what the company knows, built from the actual interactions where knowledge is created: calls, meetings, messages, and the browsing and research that happens between them. That is what makes it different from knowledge management. Knowledge management asks people to organise what they know. A memory layer captures it without asking.

Why Agentic AI Makes This Urgent

There is a reason this problem is suddenly urgent. The market is shifting toward agentic AI: systems that do not just respond to prompts but take autonomous action, making decisions, handling tasks, and operating across applications. Companies everywhere are building or buying AI agents to handle work that used to require a person in the loop. The promise is significant. But there is a catch that not enough people are talking about.

AI agents are only as good as the data they have access to.

If most of a company's knowledge was never recorded, the agent is working with a fraction of the picture. It can search the CRM, scan documents, read messages, and still miss the thing that actually matters. Because the thing that matters was said on a call that nobody summarised, in a meeting that produced no notes, in a conversation where both people assumed the other would remember.

Before an AI agent can reason about a business, it has to have access to what that business knows. And most of what the business knows was never recorded. You cannot build intelligence on top of absent memory.

The companies building company brains right now are mostly focused on connecting tools. Indexing documents. Building knowledge graphs across existing data. That is useful work. But it assumes the knowledge is already sitting somewhere, waiting to be connected. For most companies, the bigger problem is upstream. The knowledge never made it into any system.

The most sophisticated knowledge graph in the world is only as complete as its inputs. And if agentic systems are reasoning on top of an incomplete graph, they will make confident decisions based on partial information.

The companies that win will not be the ones with the best models or the most sophisticated orchestration. They will be the ones who understood that the data layer had to come first. That before you can reason, you have to remember. And before you can remember, you have to capture.

The Company Brain Starts With Individual Context

There is a lot of energy around the company brain concept. Build a system that knows everything the company knows. Connect every tool, index every document, map every relationship. It is an appealing vision.

But most people are starting in the wrong place.

A company's knowledge does not originate in documents or databases. It originates in individual interactions. A person has a call. A person visits a prospect's website. A person sits in a meeting and hears something that changes how they think about a deal. The knowledge starts with the individual, always.

If each person's context is broken, if they cannot recall what happened three months ago, if they cannot surface the right detail at the right moment, if the information they learned disappears the moment they move on to the next task, then the company's knowledge is broken too. No amount of enterprise architecture fixes that.

The company brain is a real idea. But it is a second-order problem. The first-order problem is giving each person the ability to retain and recall context from their own work. Fix that, and the organisational memory follows naturally. The knowledge that was captured passively at the individual level becomes available to the team, the department, the company. It compounds.

Try to build it from the top down, connect the tools, index everything, build the enterprise graph, and you end up with another system that is only as good as what people bothered to put into it.

Where the Company Brain Actually Gets Built

The answer is not better CRMs. It is not more fields to fill in. It is not another tool that asks people to do more admin. Every company that has tried the "just make people log things properly" approach has learned the same lesson: compliance is not capture.

The answer is ambient knowledge capture. Knowledge captured as a byproduct of work, not as a separate task. The information flows into the system without the human having to stop, switch contexts, and perform data entry.

Soda is built on this principle. It runs quietly in the background, paying attention to what is happening across your calls, your browser, and your apps, and captures the knowledge that would otherwise disappear. Not by adding another tool to the stack, but by sitting underneath the tools you already use and remembering what you would want to remember.

The company brain starts here. Not with the enterprise knowledge graph. Not with the connected tool index. With each person's context, captured without effort, retained over time, and available when it matters.

Build the memory layer for individuals, and the company brain assembles itself.