Someone leaves the company.
Maybe they retire after twenty years. Maybe they take another job. Maybe they simply decide it is time to do something different. There is a farewell lunch, their laptop gets collected, accounts are reassigned, and everyone assumes the transition will be manageable.
Then the questions start.
Why does this customer receive different terms than everyone else? Why did we stop using that vendor? Where does the number in this spreadsheet actually come from? Why does this process have an extra approval step that no longer seems necessary? What happened the last time we tried to change it?
The documents may still exist. The spreadsheet is still on the shared drive. The emails are probably sitting in an archive somewhere.
What disappeared was the context that made those things make sense.
Most businesses do not think of themselves as having a memory problem. They think they have a documentation problem, a training problem, a process problem, or occasionally a problem with one person who knows far too much.
Those are often different symptoms of the same thing.
The business learned something, sometimes at considerable expense, and never developed a reliable way to keep what it learned.
A company knows more than its files
We tend to think of business knowledge as information: policies, contracts, customer records, procedures, presentations, reports, project plans, meeting notes. Things that can be saved, searched, organized, and placed into a folder.
Those things matter, but they are only part of what an organization knows.
A great deal of useful knowledge lives in the space between what happened and why.
A contract may tell you that a customer received an unusual concession, but not that it was made after a shipment failed three days before the customer's largest event of the year. A procedure may contain an extra quality check without explaining the expensive mistake that caused someone to add it. A project folder may contain three abandoned approaches without preserving the reasoning that finally sent the team in another direction.
Experienced people accumulate thousands of these connections.
They know which rules can bend and which ones absolutely cannot. They remember the customer who cares deeply about something that never appears in the CRM. They know why the seemingly obvious solution was already tried four years ago and failed. They recognize the strange spreadsheet as a workaround for two systems that have never talked to each other properly.
This context is not trivia surrounding the work. It is often the part that makes the work intelligent.
When it remains attached to a person instead of the organization, a company can appear well documented while still being surprisingly fragile.
The documentation problem was never a lack of places to put things
We have spent decades trying to solve this.
Shared drives became intranets. Intranets became wikis and knowledge bases. Teams built carefully organized systems in tools like Notion. Companies created naming conventions, templates, folder structures, standard operating procedures, and increasingly elaborate rules about where information should live.
Some of these systems are excellent.
The problem is usually not the place where the knowledge goes. It is getting people to put it there.
Traditional knowledge management asks people to do something that competes directly with the work itself. Finish the meeting, then document the decision. Solve the customer problem, then update the CRM. Change the process, then revise the SOP. Complete the project, then record what everyone learned while the team is already moving toward the next deadline.
The system works beautifully if people continuously maintain it.
Then people get busy.
A decision gets captured in a Slack thread instead. Someone says, "I'll update that later." Notes pile up. Documentation drifts away from reality. Six months later, somebody finds the old version and assumes it is still correct.
The theory was never particularly difficult. The maintenance was.
AI changes that equation
This is one of the places where artificial intelligence becomes much more interesting than simply generating another document.
Imagine using AI while the work is actually happening.
You ask a question. You work through a difficult decision. You compare options. You create a proposal or analyze a customer problem. Somewhere in that work, something emerges that the organization should not have to rediscover later.
A well-designed system can recognize that.
It can draft the note, capture the decision and its reasoning, place it with the appropriate project or subject, connect it to related knowledge, and make it available when that context becomes relevant again.
The human contribution can become remarkably small: yes, that matters. Keep it.
Then go have a snack.
That sounds almost trivial, but it removes one of the largest barriers that has kept knowledge-management systems from fulfilling their promise.
The gap between having a useful insight and preserving it has become dramatically smaller.
Instead of asking people to maintain an elaborate second job called "document everything you know," we can increasingly design systems in which useful memory develops as a consequence of doing the work.
That is a fundamentally different proposition.
We are already beginning to see versions of this emerge inside sophisticated organizations. In episode 231 of The Artificial Intelligence Show, "How Zapier Is Building an AI Marketing Brain", Mike Kaput spoke with Zapier's Chief Marketing and AI Transformation Officer Dan Slagen about the company's work on an AI-powered marketing brain: a shared system intended to keep current context about Zapier's market, customers, competitors, campaigns, goals, and brand available to the AI tools and agents helping the marketing team work.
What I find most interesting is not that Zapier also chose the word brain. It is what the system requires from the organization around it. Slagen described changing how the team communicates and works in public so useful knowledge does not disappear inside private conversations the brain cannot see. The technology can help preserve and connect context, but the organization still has to decide which context becomes available in the first place.
That is the shift. The old knowledge-management problem was largely about getting people to maintain the repository after the work was done. AI gives us a chance to make capture, organization, and connection part of the work itself.
The model can know you while the business still cannot remember itself
AI has also given us a strange new experience of memory.
A model can learn how you like something written, retain preferences, recognize recurring projects, work across previous conversations, and maintain enough continuity that returning to it begins to feel less like opening software and more like continuing a conversation.
That can be incredibly useful.
It can also obscure an important distinction.
An executive can have hundreds of thoughtful conversations with an AI about strategy, customers, employees, decisions, and problems inside the company. A manager can use another model every day to reason through projects. Someone else may create increasingly sophisticated assistants around their own work.
Each person can become dramatically more capable while the organization remains just as forgetful as it was before.
The model can know you while the business still cannot remember itself.
If useful context remains scattered across individual accounts, conversations, inboxes, spreadsheets, and people's heads, we have not solved the organizational-memory problem. We have given it newer places to hide.
Memory is not everything that happened
The answer is not to save everything.
Recording every meeting, conversation, draft, passing thought, and AI interaction would create an enormous archive without necessarily creating a useful memory.
A company needs judgment about what deserves to persist.
Important decisions and the reasoning behind them. Commitments someone will need to honor later. Exceptions and why they exist. Lessons from projects that succeeded or failed. Customer context that changes how a relationship should be handled. Processes that evolved for reasons worth preserving.
The objective is not perfect recall. It is to stop repeatedly paying for knowledge the organization has already earned.
That distinction becomes even more important as AI makes capturing information nearly effortless. We can produce more summaries, transcripts, notes, and documents than any human being could reasonably consume.
Saving more is not the breakthrough. Creating useful continuity is.
The memory should belong to the business
The AI systems we use today will change.
Models will improve. Products will come and go. Interfaces will change. The tool that feels indispensable today may not be the one we want to use several years from now.
That is normal.
The accumulated understanding of the business should not disappear with it.
A company's important knowledge should be able to survive the employee who created it, the application that stored it, and the model that helped make sense of it. The intelligence can change. The memory should endure.
This is the problem we built Spark Evolution's Executive Second Brain around.
Not a giant archive where everything a company has ever produced gets dumped into one place, and not an attempt to replace the people whose experience makes the business valuable.
The goal is to create a durable layer of useful context the business owns, then let AI help capture it, organize it, connect it, and bring it back when it matters.
Because every organization will continue learning difficult lessons.
It should not have to keep learning the same ones.
Kurt Milligan is co-founder and CEO of Spark Evolution. If something here matches a problem you are carrying, start a conversation.