What happens to institutional memory?

Institutional memory splits into two parts that behave differently, and the question usually collapses them. The recorded part becomes far easier to retrieve. The unrecorded part, which includes why decisions were taken, what was tried and abandoned, and how things are actually done, moves through people and is untouched by making documents searchable.

So an organisation can arrive at better retrieval and weaker memory at the same time, and the first is measurable while the second is not.

Retrieval and retention#

Polanyi's argument is the reason the two come apart. We can know more than we can tell. A large part of what an organisation knows was never written down, not through negligence but because it could not be articulated. The Tacit Dimension (1966).

Putting the archive into a system that can answer questions about it is a genuine improvement to the first part. It changes nothing about the second, because the second was never in the archive.

The failure mode is specific and common: an organisation concludes from the accessibility of its records that its knowledge is secure. What it has secured is the residue.

What the unrecorded part actually contains#

Worth being concrete, because "tacit knowledge" is used so loosely that it stops meaning anything.

The reasons a decision was taken, as distinct from the decision. The options considered and rejected, which almost never survive into the document. Which parts of the written process are load-bearing and which are ceremonial. Who to ask. What went wrong last time this was attempted. How much a given estimate can be trusted, and by whom.

None of that is in the file. Most of it is in someone's head, and it transfers when people work together on something that matters, which is Nonaka and Takeuchi's account of how the tacit part moves at all. The Knowledge-Creating Company (1995). Their model has been criticised for treating the conversion from tacit to explicit as more tractable than Polanyi allowed, and the criticism holds, but the observation about mechanism survives it.

The knowledge-sharing question, which is open#

There is an obvious hypothesis here and the evidence does not yet support stating it as fact.

If people can get task information from a system instead of from a colleague, some of the instrumental relationships through which workplace knowledge travelled will weaken. That is plausible and it is not established. Early work points towards rerouting rather than reduction, with task-oriented ties weakening while other social connection does not, which is a more interesting result than a straightforward collapse and rests on very little.

What is worth noticing is which part is most substitutable. Asking a colleague a question you could have looked up was never only about the answer. The context that came back unbidden, the correction of the question itself, the knowledge that you were the sort of person who asks: those were by-products of an inefficiency. Remove the inefficiency and the by-products go with it.

That is a mechanism, not a finding, and this estate lists the question as open rather than answering it.

What protects it#

Three practices, none of which competes with a searchable archive and none of which is replaced by one.

Handovers with an overlap rather than a document. The document is worth writing. The overlap is where the rest transfers.

Decision records that state what was rejected. The cheapest way to move a large amount of otherwise unrecoverable reasoning into the written part, and almost nobody does it, because the rejected options feel like clutter at the moment of writing and are the entire value a year later.

Keeping the routine of asking someone who was there. Argyris and Schon's point applies: an organisation optimising a measured loop will remove the unmeasured contact inside it without ever deciding to. Organizational Learning (1978).

What this page does not establish#

No study shows that AI adoption degrades institutional memory. The mechanism is available, the distinction between retrieval and retention is well founded in the tacit-knowledge literature, and the measurement has not been done.

The underlying sources are philosophical and case-based rather than experimental. Polanyi was not testing anything, and Nonaka and Takeuchi built from Japanese manufacturing cases whose generalisation is contested.

There is also a real counter-position. Some of what organisations call institutional memory is accumulated habit that survives because nobody has re-examined it, and making the reasoning retrievable can expose that rather than preserve it. Better retrieval is not automatically a loss dressed as a gain.

Key sources

• Polanyi, M. (1966). The Tacit Dimension. University of Chicago Press. In the essential works.
• Nonaka, I. and Takeuchi, H. (1995). The Knowledge-Creating Company. Oxford University Press. In the essential works.
• Argyris, C. and Schon, D. A. (1978). Organizational Learning: A Theory of Action Perspective. Addison-Wesley. In the essential works.
• Nelson, R. R. and Winter, S. G. (1982). An Evolutionary Theory of Economic Change. Belknap Press. In the essential works.

Related SuperSkills research#

On what the organisation holds, organisational capability. On the part that cannot be written down, tacit knowledge. On holding on to the people who carry it, keeping expertise in an organisation. On what the loss accumulates as, capability debt. On measuring it, assessing capability rather than output.

About this research#

Rahim Hirji is the author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. The tacit dimension is Polanyi's and the knowledge-creation model is Nonaka and Takeuchi's. Neither is a SuperSkills coinage.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.