Every business has a question that only one person can answer. How do we invoice that client who pays through a portal? Which supplier do we use when the usual one is out? Where's the login for the old booking system, and do we still need it?
Usually the answer is "ask Dana." That works fine until Dana is on holiday, or busy, or leaves.
Knowledge management is the work of making sure the answer doesn't depend on Dana. This post covers what knowledge management means, the difference between tacit and explicit knowledge, the process most teams follow, what a knowledge manager does, and why so many internal wikis end up as museums. The last part covers AI knowledge management. We make Sanaf, an AI employee that can take on this job, so we have an interest in that section. Everything before it holds whoever does the work.
The short answer
Knowledge management is how a company captures what its people know, organizes it so others can find it, keeps it accurate as things change, and gets it to the person who needs it at the moment they need it.
In practice that means procedures, how-to guides, policies, answers to common internal questions, customer history, supplier details and the reasoning behind decisions. A knowledge management system is whatever holds all of that: a wiki, a shared drive, a help center, a set of pinned chat messages, or more often an untidy mix of all four.
The definition sounds like filing. The hard part is everything after the filing, which is keeping the knowledge true and getting people to use it.
Why knowledge management matters more than it sounds
Most owners don't think they have a knowledge management problem. They think they have a training problem, or a "people keep asking me the same thing" problem, or a "we did this wrong again" problem. Underneath, those are usually one problem.
Here's what it looks like day to day:
- A new starter spends their first month shadowing someone, because nothing is written down well enough to learn from.
- The same question gets asked in the team chat every few weeks, and answered slightly differently each time.
- A procedure exists, but it describes how things worked two years ago, and following it causes a mistake.
- Someone experienced leaves and takes the only copy of how a key process works with them.
- The owner can't take a week off without their phone ringing.
That last one is the tell. When the business can't run without one person in the room, the business doesn't really own its own processes. That person does. That's what people mean by institutional knowledge: the shared understanding of how things actually get done here, which often lives nowhere except in heads.
Tacit vs explicit knowledge
The most useful idea in knowledge management is the split between two kinds of knowledge.
Explicit knowledge is what can be written down and handed over. A price list, a refund policy, the steps for setting up a new client account, the supplier's phone number. If you can write it in a document and someone else can follow it, it's explicit.
Tacit knowledge is what people know but find hard to put into words. Which customers need a phone call instead of an email. How to tell a supplier's "we'll try" from a real yes. The feel for when a job quote is too low. The philosopher Michael Polanyi put it as "we can know more than we can tell," and the phrase stuck because everyone recognizes it.
| Explicit knowledge | Tacit knowledge | |
|---|---|---|
| What it is | Facts, steps and rules that can be written down | Judgment, experience and know-how that's hard to express |
| Examples | Price list, onboarding checklist, refund policy | When to bend the policy, which client needs extra care |
| Where it lives | Documents, wikis, drives, help centers | People's heads, and the stories they tell |
| How it's shared | Reading, searching, linking | Working alongside someone, asking, watching |
| What goes wrong | It goes out of date | It walks out the door |
Good knowledge management works on both. Much of the job is turning tacit knowledge into explicit knowledge where you can: asking the experienced person how they decide, writing the reasoning down, and checking with them that the document says what they meant. The rest is accepting that some knowledge will stay tacit and making it easy to find the person who has it.
The best-known model of this movement comes from Nonaka and Takeuchi's SECI model, which describes knowledge moving between tacit and explicit forms through socialization, externalization, combination and internalization. You don't need the vocabulary to run a good process, but it's where most knowledge management frameworks start.
The knowledge management process
Most knowledge management frameworks describe a cycle with some version of these steps, though the names change from one framework to the next.
- Capture. Get knowledge out of heads, emails and chat threads and into a form someone else can use. Interviewing the person who does the job, recording how a task is done, saving a good answer the first time it's given.
- Organize. Put it where people will look for it, with a name they would search for. A procedure called "Process v3 FINAL" in someone's personal drive has been captured but not managed.
- Share. Get the knowledge to the person at the moment they need it. In most companies this happens in chat, by someone asking a question, so the answer has to be reachable from there.
- Use and learn. People follow the procedure and find where it's unclear or wrong. That feedback is the most valuable input in the whole cycle, and the one most often lost.
- Update or retire. Fix what changed, merge duplicates, archive what no longer applies. This is the step that never gets done.
Step five is why the knowledge management cycle is a cycle and not a project. A knowledge base built once and left alone starts dying on the day it's finished, because the business keeps changing and the documents don't.
What does a knowledge manager do?
In large companies, a knowledge manager is a dedicated role. In most businesses nobody holds the title, and the work gets done in scraps by whoever is most organized or most annoyed. Either way, the duties look similar:
- Maintains company knowledge. Decides what should be written down and makes sure it gets written.
- Organizes standard operating procedures. Keeps SOPs consistent enough to follow, with an owner for each one.
- Updates documentation. Changes the document when the process changes, before someone follows the old version.
- Answers internal questions. Points people to the right document, or answers and then writes the answer down.
- Finds information across systems. Knows that the real answer is often in an email thread or a chat message, not the official wiki.
- Identifies outdated documentation. Spots what's gone stale and flags it to whoever owns it.
The skills behind those duties are less glamorous than they sound. Good knowledge managers are curious enough to ask "why do we do it this way?", patient enough to interview busy people, plain enough in their writing that a new starter can follow the result, and stubborn enough to keep updating documents nobody thanks them for.
Knowledge sharing and knowledge transfer
Two related terms come up often.
Knowledge sharing is the everyday flow of knowledge between people: asking a colleague, answering in a channel, explaining something in a meeting. Most of a company's knowledge sharing happens in conversation, which is fast and friendly and leaves almost nothing behind.
Knowledge transfer is the deliberate handover of knowledge from one person or team to another, usually because something is changing. Someone is leaving, a role is being split, a process is moving to a new team. The classic failure is the two-week notice period where the departing person writes a long handover document that nobody reads until something breaks three months later.
The fix for both is the same habit: when an answer is given in conversation, it becomes a document, or updates one. Do that consistently and knowledge transfer stops being an emergency, because most of what the leaving person knows was already captured, one question at a time.
Why internal knowledge bases go stale
Almost every company has a wiki that was accurate in March. The reasons are nearly always the same.
Nobody owns maintenance. Writing the first version feels like progress, so somebody does it. Keeping it current feels like housekeeping, so nobody does.
Staleness is invisible. A wrong document looks exactly like a right one. Nobody notices a procedure is out of date until they follow it and something goes wrong, and by then the damage is done.
The answers live somewhere else. People ask in chat because it's faster than searching the wiki. The answer gets given in chat. The wiki never hears about it.
Search is bad, so trust drops. People search, get three conflicting results, and go back to asking a person. Every time that happens, the wiki gets less traffic and less maintenance, and the cycle tightens.
Duplicates multiply. Two people write the same procedure in two places a year apart. Now there are two versions, and no one knows which one is current.
None of this is fixed by buying a better knowledge management system. The tool matters less than whether someone is doing step five of the cycle every week.
A knowledge management strategy you can actually run
A knowledge management strategy doesn't need to be a 40-page plan. For most businesses, five decisions cover it:
- Pick one home. Choose where official knowledge lives. It can be Notion, Google Drive, SharePoint or anything else, but pick one, and link to it from chat.
- Start with the questions people ask. Don't try to document everything. Look at the last month of internal questions and write down the answers to the ten most common ones. That's your first knowledge base.
- Give every document an owner and a review date. The owner is the person who'd notice if it went wrong. The review date can be a rough one, every three or six months.
- Make "write it down" part of answering. When someone answers a question that isn't documented, the answer becomes a document or updates one, the same day.
- Retire ruthlessly. Archive anything nobody has opened in a year and nobody can vouch for. A smaller, true knowledge base beats a large one full of maybes.
Notice that only one of those is about tools. Knowledge management best practices are mostly about habits and ownership, which is also why they're hard to keep up when everyone is busy.
What is AI knowledge management?
AI knowledge management uses AI to do parts of the cycle above: answering internal questions from your documents, searching across several systems at once, drafting procedures from conversations, and flagging documents that look out of date.
The appeal is obvious. The work that never gets staffed, maintenance and answering the same question for the twelfth time, is exactly the kind of steady, repetitive work AI handles well. An AI that sits in the team chat can answer where people actually ask, and turn those answers into documentation without anyone having to remember.
The risk is just as obvious, and it's the one that matters most. AI models can produce confident, fluent answers that are wrong, which is usually called AI hallucination. For a public chatbot, a wrong answer is embarrassing. For internal knowledge, it's worse, because a confident wrong answer about your own process gets acted on. Nobody double-checks the refund policy when the company's own assistant just stated it.
So the most important rule for an AI knowledge manager is simple: it never invents an answer. A good one:
- Answers from what is actually written down, and shows where the answer came from so anyone can check.
- Says "I don't know" when the documentation doesn't cover the question, and brings in the person who does know.
- Turns that person's answer into documentation, so the gap closes once instead of being rediscovered next month.
- Flags documents that contradict each other or reference tools you stopped using, and asks the owner instead of quietly rewriting them, because deciding what's current is a judgment.
- Searches the places knowledge actually lives (drive, chat, email, wiki) instead of only the official one.
An AI that can't say "I don't know" shouldn't be anywhere near your internal knowledge. When you test any AI knowledge base tool, ask it a question about your business that you know isn't documented. What it does next tells you most of what you need to know.
What AI doesn't fix is ownership. Someone still has to decide what the right process is. AI can notice that two procedures disagree. It can't decide which one the business should follow.
The AI knowledge management page goes into how that job is set up, and the same habit of writing the answer down after the first time shows up in IT support and HR.
Where Sanaf fits
Sanaf is an AI employee. It works in your team's Slack or Microsoft Teams and inside the apps you connect, such as Notion, Google Drive, Gmail and Outlook. You give it jobs from a roster, and knowledge management is one of them: maintaining company knowledge, organizing standard operating procedures, updating documentation, answering internal questions, finding information across the systems you connect and identifying documentation that has gone stale. The AI employee roster lists the other jobs.
The rule it works under is the one above. When the documentation doesn't cover a question, it says so and brings in the person who knows. It asks before anything it can't undo, and every job leaves a work log the team can read in the channel. The security page explains how access and approvals work.
On price, you start with $120 of free credit and no card, and the free credit lasts 30 days. After that you buy credits from $20 when you need them, and credit you buy never expires. You pay for the work, not per seat. The pricing page has the details, and hiring an AI employee shows what the first day looks like.
If you want a quick test, take the three questions your team asks most often and the one that only Dana can answer. Ask all four. Three good answers with sources and one honest "I don't know, let me ask Dana" is what you're looking for.
Questions people ask about knowledge management
What is knowledge management?
Knowledge management is how a company captures what its people know, organizes it so others can find it, keeps it accurate as things change, and shares it with the people who need it. It covers procedures, policies, how-to guides, answers to common questions and the reasoning behind decisions.
What is the definition of knowledge management in simple terms?
In simple terms, knowledge management means making sure the answer to "how do we do this here?" is written down, easy to find and still true, so the business doesn't depend on one person's memory.
What is the difference between tacit and explicit knowledge?
Explicit knowledge can be written down and followed by someone else, such as a price list or a checklist. Tacit knowledge is experience and judgment that's hard to put into words, such as knowing which client needs extra care. Knowledge management tries to turn tacit knowledge into explicit knowledge where possible, and to make the people who hold the rest easy to find.
What is institutional knowledge?
Institutional knowledge is the shared understanding of how things actually get done inside a company: the processes, history, relationships and unwritten rules. It's often undocumented and held by long-serving people, which is why it's at risk when they leave.
What are the steps of the knowledge management process?
Most knowledge management cycles include capturing knowledge, organizing it, sharing it, using it and learning from that use, and then updating or retiring it. The update step is the one most often skipped, and it's why knowledge bases go out of date.
What does a knowledge manager do?
A knowledge manager maintains company knowledge, organizes standard operating procedures, updates documentation, answers internal questions, finds information across company systems and identifies outdated documents. In many businesses there's no one with the title, and the work is shared or skipped.
What is a knowledge management system?
A knowledge management system is the tool or set of tools that holds a company's knowledge, such as a wiki, a shared drive, an internal knowledge base or a help center. The system matters less than whether someone keeps its contents current.
What is the difference between knowledge sharing and knowledge transfer?
Knowledge sharing is the everyday flow of knowledge between colleagues, mostly through conversation. Knowledge transfer is a deliberate handover from one person or team to another, often because someone is leaving or a process is moving.
What is AI knowledge management?
AI knowledge management uses AI to answer internal questions from company documents, search across several systems at once, draft documentation from conversations and flag documents that look outdated. A trustworthy one answers only from what's written, shows its sources and says "I don't know" when the documentation doesn't cover a question.
Can AI replace a knowledge manager?
AI can carry much of the repetitive work: answering common questions, searching across systems, drafting updates and spotting stale documents. People still decide what the correct process is, resolve contradictions and hold the tacit knowledge AI can't see. Most teams end up with AI doing the upkeep and people owning the decisions.