timer://capture-institutional-knowledge
How to capture institutional knowledge.
Every organization loses knowledge, and almost every organization tries to fix it with documentation drives, handover templates and exit interviews. Here is why those fail, and what works instead.
01 Why the usual methods fail
All four fail the same way. They ask a person to do the remembering.
Institutional knowledge is the accumulated understanding of how an organization actually works: the reasoning behind decisions, the exceptions to documented processes, the history of relationships, the judgment calls experienced people make without being asked. It is held in people rather than systems, which is why it leaves when they do. The standard responses all treat that as a discipline problem, and it is not.
Documentation drives
Compete directly with the work people are measured on, so they finish at roughly the moment the pressure is removed. What survives is the easy half: the parts already obvious enough to write down quickly.
Handover documents
Ask someone to articulate, in two weeks, context accumulated over years. Most of it is invisible to the person holding it, because expertise feels like common sense from the inside.
Exit interviews
Arrive after the person has mentally left, and are structured around how they felt about working somewhere rather than what only they know.
Wikis and shared drives
Capture what somebody deliberately sat down to write. Institutional knowledge is, by definition, the part nobody sat down to write, and a wiki has no way of knowing what is missing from it.
The common failure is structural. Each method depends on a person choosing to record something, at a moment when recording it competes with everything else. Capture that depends on intention degrades to whatever is left over after the real work, which is why organizations that have run all four still cannot explain why a policy exists.
02 The method
Six steps, in order. The order matters more than the tooling.
Decide what actually needs keeping
Not everything is worth capturing, and trying to capture everything is why documentation programmes collapse. The unit worth keeping is the decision: what was chosen, on what basis, and what it committed the organization to. Documents describe the state of the world at one moment. Decisions explain why the world is the way it is, and they are what nobody can reconstruct later.
Capture at the moment of work, not the moment of exit
Knowledge is cheap to record while the work is happening and effectively impossible to recover afterwards. Anything that asks a person to stop and write up what they know is competing with their actual job, and loses. Capture has to be a property of the systems people already work in, so the record accumulates as a side effect of doing the work rather than as a separate task nobody has time for.
Record the reasoning, not just the outcome
Most systems already store outcomes: the contract, the approval, the final number. Almost none store why. The reasoning is what makes a record reusable, because it is what lets a successor tell whether the original constraint still applies. A rule everyone follows and nobody can explain is what an organization is left with when only outcomes were kept.
Give it one home the whole organization reads from
Knowledge split across a wiki, a drive, a chat history and six departmental tools is not captured, it is scattered, and scattered knowledge fails in the same way undocumented knowledge does: nobody can find it when it matters. One memory layer that every function and every AI system reads from and writes to is what turns a pile of records into something an organization can actually inherit.
Govern who can read what, and be able to prove it
Memory that accumulates without governance becomes a liability rather than an asset, particularly in regulated work. Access has to be scoped, the record has to be tamper-evident, and it has to be possible to show after the fact who saw what and who decided what. Governance is what makes it safe to capture more rather than less.
Own it independently of any vendor
Knowledge captured inside a tool belongs to the organization only for as long as it keeps paying for that tool on those terms. If the record cannot leave, it was never fully captured. The test is simple: if the vendor changed its pricing, its terms or its roadmap tomorrow, would the organization still hold its own history in a form it could move?
03 What to capture
Five things, per decision.
A useful record of a decision has five parts. Most organizations already keep the first and the last, and lose the three in the middle, which are the three that make the record worth anything to a successor.
The decision
What was chosen, stated plainly enough that someone outside the room can tell what changed.
The reasoning
Why that option and not the obvious alternative. This is the part that lets a successor tell whether the original constraint still holds.
The constraint
What the organization was not free to do: the regulation, the contract, the commitment made to someone two years earlier.
The person
Who decided, and who reviewed. Without a name, accountability cannot be reconstructed and the record cannot be used as evidence.
The outcome
What actually happened, recorded against the decision rather than in a separate system, so the organization can learn from the gap between the two.
Held together and in order, these five are what Timer calls decision lineage: the atomic unit of organizational memory, and the difference between an archive and something an organization can actually learn from.
04 Questions