Knowledge Systems as Executive Infrastructure

The Organisation That Cannot Answer Itself

Almost every organisation of a certain size believes it possesses knowledge. It has policies and playbooks, shared drives and case files, veterans with three decades in the building and graduates who arrived last month. It has spent, over the past twenty years, remarkable sums on systems whose stated purpose was to make what it knows available to itself. And yet, when a specific question needs answering on a Tuesday afternoon, the same organisation will often find that the answer lives in the head of one person, that this person is on leave, and that the last time the question was resolved no one recorded how.

The first essay in this series was concerned with the question of whose judgement shapes an important technology decision. That essay treated independence as the rarest quality in the room. This essay is concerned with a quieter question that sits underneath the first one. Even where independent judgement is present, and even where it is welcomed, what are the organisational conditions under which such judgement can be exercised at all? A well advised board still requires an organisation capable of receiving the advice, remembering it, and acting on it a year later when the person who gave it has moved on.

The subject of this essay is therefore not artificial intelligence, though artificial intelligence will appear in it repeatedly. The subject is the older and less fashionable matter of organisational knowledge, and the reasons why so few institutions actually possess it, in spite of owning very large quantities of the raw material from which it might, in principle, be made.

Information, Knowledge, Expertise

It helps to be careful about words. Information is not knowledge, and knowledge is not the same thing as expertise. An organisation that has stored a document has not, by that act, learned anything. An organisation that employs an expert has not, by that act, retained what the expert understands. Knowledge, in the sense that matters for institutions, is the capacity to act coherently on the basis of what has been learned before. It is a property of the organisation as a whole, not of any particular file, system or individual within it. It emerges when distributed experience is turned, again and again, into coordinated action, and it decays whenever that transformation stops happening.

This is worth stating plainly because a great deal of what passes for knowledge management, in ordinary corporate life, is the management of documents. The two activities look similar and are frequently confused. Documents are traces of thought. Knowledge is the capacity of an institution to think again in the same way when it needs to, without asking each new person to reconstruct the reasoning from first principles. A firm with excellent document management and poor institutional memory is more common than one might expect. It is also more common than the firms in question tend to realise.

Twenty Years of Accumulation

The past two decades have been, by any reasonable measure, an era of extraordinary digitisation. Contracts moved from filing cabinets to shared drives, then to contract lifecycle systems. Customer correspondence moved from letters to email to platforms whose purpose was to make every prior interaction retrievable. Enterprise resource planning became the connective tissue of finance and operations. Collaboration tools multiplied. SharePoint arrived, then Confluence, then Notion, then a fresh generation of workspace products that promised, each in turn, to be the place where the organisation would finally think together.

The volume of stored information inside a typical mid sized business today is many multiples of what it was in the early two thousands. The number of systems in which that information lives has grown at a similar pace. What has not grown, in any comparable way, is the ease with which an ordinary employee can answer an ordinary question about the organisation's own past. Studies of professional work continue to show that experienced staff spend a striking share of their week searching, asking colleagues, and reconstructing conclusions that were reached, and often documented, somewhere in the estate. The material is present. The organisation, in the sense that matters, cannot get to it.

The Mirror the Model Holds Up

The reason is not primarily technical. It is that accumulation and organisation are different activities, and the past twenty years have been overwhelmingly devoted to the first. Each new system added a further place in which information could be stored. Very few added a discipline for deciding what belonged there, in what form, under whose stewardship, and for how long. The result is an estate in which almost everything exists and almost nothing can be relied upon. Confluence pages contradict SharePoint folders, which contradict a set of PDFs on a private drive, which contradict what the head of a business unit remembers being agreed at the time.

Into this environment arrives, with considerable fanfare, artificial intelligence. It is tempting to describe it as the answer to the problem. That would be a mistake. Artificial intelligence, in its current forms, is not a solution to fragmented knowledge. It is a mirror held up to it. A model can only reason with what has been made available to it in a form it can use. When an organisation asks a well constructed system a serious question about its own operations, and the answer comes back thin, contradictory or confidently wrong, the model is not usually the problem. The problem is that the organisation had never, before that moment, been asked to state clearly what it knew.

This is the useful function that artificial intelligence performs for the institutions willing to look. It exposes, quickly and without malice, the architecture of the organisation's own memory. Firms that have quietly maintained coherent bodies of decisions, case notes, precedents and standards over many years find that their systems become sharply more capable once given proper access. Firms that have relied on the memory of a few senior figures find that their systems remain oddly disappointing regardless of which model is chosen or how large the licence fee. The difference is not in the technology. It is in what the technology has been given to work with.

Memory as Infrastructure

It is worth distinguishing three things that are often blurred together in ordinary conversation. The first is individual expertise, which resides in a particular person and leaves the building when that person does. The second is collective experience, which is what a team has lived through together, and which persists as long as enough of the team remains. The third is institutional memory, which is the property of the organisation itself, and which survives the departure of any given individual or team because it has been captured, structured and made retrievable in a form the next generation can use.

Most organisations are strong on the first, patchy on the second and thin on the third. They employ excellent people. They have teams with real shared history. And when a senior partner retires, or a lead engineer moves to a competitor, or a business unit is reorganised, the institution discovers that a substantial portion of what it thought it knew has quietly walked out. What remains behind, in the systems, is often the record of what was done, without the record of why. When the people go, the reasoning goes with them.

Institutional memory, understood in this way, is best thought of as infrastructure. It has more in common with roads, electricity supply, drainage and double entry bookkeeping than it does with any particular software product. Infrastructure is the class of thing that is almost invisible when it works. Nobody in a functioning city thinks about the water supply while making a cup of tea. The moment the supply fails, however, everything above it fails as well, and the failure is total rather than partial. Institutional memory has the same character. Its presence is felt as ordinary competence. Its absence is felt as the strange and expensive slowness that afflicts organisations which know, in some abstract sense, a great deal, but cannot bring what they know to bear on the question in front of them.

Because it is infrastructure, institutional memory is an executive concern rather than a technology concern. This is a distinction that matters. Roads are not the responsibility of the paving department. Electricity is not the responsibility of the wiring team. These are civic questions, decided at the level of the institution as a whole, because the whole institution depends on them. Institutional memory belongs in the same category. When it is delegated, as it often is, to a knowledge management function two levels down in the organisation, or to a corner of the technology group with no standing at the executive table, the outcome is predictable. The function does its work honourably and the organisation continues to lose its memory at the top.

What This Looks Like in Practice

It is easier to see this through examples than through argument, so it may help to describe a few composite situations that will be recognisable to anyone who has spent time inside serious institutions. None of these is a single real firm. All of them are true.

A well regarded law firm finds itself, over the course of a year, working on four separate matters that turn on the same narrow question of contractual interpretation. Each matter is staffed by a different team. Each team reaches, after several days of research, substantially the same conclusion. None of them is aware of the other three. The firm's precedent system contains, in principle, the earlier memoranda. In practice, the memoranda are filed under client names rather than under the question they answer, and the partners involved have long since moved to other matters. The firm bills for the work four times. It has learned the question once.

A tax advisory practice with a strong reputation in a particular jurisdiction discovers, during a routine review, that three of its senior advisors hold subtly different views on the treatment of a common structure. Each view is defensible. Each has been given to clients over the past decade. No one at the firm can say with confidence which view is now considered house position, because no such position was ever formally recorded. The founding partner, who would have known, retired four years ago. What the firm possesses is expertise. What it lacks is a memory of its own opinions.

A manufacturing company loses, within eighteen months, three lead engineers to retirement. Each had spent thirty years with the business. Each held, in their head, the reasons why certain unusual choices had been made in the design of the current generation of products, why particular suppliers had been trusted, why particular tolerances had been set the way they were. The drawings remain. The reasoning does not. The next generation of engineers, competent and well trained, will spend the following decade rediscovering, at considerable expense, things the firm already knew.

A software company of some maturity finds itself, every eighteen to twenty four months, having the same architectural debate. Should the platform move toward a more modular structure? Should certain services be extracted? Should the data layer be reworked? Each round of the debate is thorough, well argued, and largely unaware of the previous rounds. The company has no shortage of talented engineers. It has no institutional memory of its own architectural reasoning, and so it pays, again and again, for the intellectual work of arriving at the same conclusions.

A private equity firm completes the acquisition of a mid sized industrial business. The diligence has been careful. The commercial thesis is sound. Six months into the hold period, the new owners discover that a significant portion of the operational know how, the informal understandings with key customers, the tacit rules by which the plant is actually run, exists inside the heads of perhaps a dozen employees. None of these employees is under any particular obligation to stay. The value of the asset, on close inspection, turns out to have been resting on a foundation the previous owners had never troubled to make explicit.

None of these situations is dramatic. None involves misconduct or incompetence. Each is the ordinary consequence of treating knowledge as a byproduct of work rather than as a thing worth maintaining in its own right. And each becomes materially more expensive in a period when the tools capable of acting on well maintained institutional knowledge are, for the first time, becoming widely available.

The New Economics of Memory

This is the point at which artificial intelligence stops being a background detail and becomes a strategic accelerant. Not because the models themselves are the source of advantage. They will continue to improve, they will be widely available, and the differences between them at the frontier will narrow rather than widen over time. The source of advantage is what an organisation is able to place behind the model. A firm whose institutional memory is coherent, current and trusted can put a competent system in front of a client, a partner or a new employee and expect it to reason as the firm reasons. A firm whose institutional memory is fragmented can spend a great deal on the same technology and receive, in return, a more articulate version of its own confusion.

This has a direct consequence for competitive dynamics. In previous cycles of enterprise technology, the differences between firms in the same industry were often smaller than they appeared, because most firms bought the same systems from the same vendors and configured them in broadly similar ways. The competitive question was one of execution. In the emerging cycle, the differences will be larger, because the value of the technology depends heavily on the quality of what each firm has quietly built up inside itself over the preceding decades. Two firms buying the same model, running on the same infrastructure, will get materially different results, and the gap will depend less on the software than on the state of the institution using it.

For the executive team of a serious organisation, this reframes the question. The interesting question is no longer which model to standardise on, or which platform to adopt across the group. Those are real questions, but they are second order. The first order question is whether the organisation has ever treated its own memory as a capability worth investing in on its own terms, with executive sponsorship, sustained funding, clear ownership and the same seriousness ordinarily reserved for finance, legal or safety. In most institutions, the honest answer is that it has not. The work has been done, where it has been done at all, by dedicated people operating below the level at which the resources or the authority to do it properly are held.

There is nothing exotic about the remedy. It looks, in practice, a great deal like the ordinary work of running a serious institution well. It involves deciding what the organisation needs to remember, at what level of detail, and for how long. It involves appointing people whose job is the stewardship of that memory rather than the production of new material to add to it. It involves the slow, unfashionable discipline of writing down not only what was done but why, in a form that will still be legible to a colleague five years from now. It involves accepting that a portion of every senior person's time is properly spent depositing what they know back into the institution rather than spending it on the next matter.

None of this is new. It is the counsel that the more thoughtful writers on organisations have offered for the better part of a century. It has been ignored, mostly, because the cost of ignoring it was diffuse and the benefit of attending to it was slow. What has changed is that the cost is no longer diffuse. The organisations that have quietly done this work over the past twenty years are about to discover that they built, without quite realising it, the substrate on which the next generation of tools operates. The organisations that did not are about to discover the same thing from the other side.

The Older Discipline

It is tempting, at the end of an essay of this kind, to offer a set of recommendations. That temptation is worth resisting. Recommendations imply that the problem is a matter of following the right steps in the right order, and the problem, as described here, is not of that character. It is a problem of institutional seriousness. It is answered by the decision, taken at the level where such decisions can be taken, to treat organisational knowledge as executive infrastructure and to fund it accordingly, over years, without expecting the returns to be visible in the current quarter.

Every generation of executives has been offered, in its time, a technology described as sufficient to compensate for the imperfections of the organisation that adopted it. It was so with the mainframe, with the client server era, with the web, with enterprise resource planning, with mobile, with cloud. In each case, the technology did what technologies do. It amplified whatever organisational habits it found already in place. Well organised institutions became more capable. Poorly organised ones became more expensively confused.

There is little reason to expect the current technology to behave differently. It is a more powerful amplifier than any of its predecessors, and the discipline that governs what it amplifies is the same quiet discipline that has always distinguished the institutions worth taking seriously from the ones that are only large. The organisations that will do well in the coming decade are the ones whose leaders understand that the interesting work has never been the choice of tool. The interesting work is the patient construction, over long periods, of an institution that knows what it knows, and remembers why.

The remaining question in this short sequence is what happens at the moment such an institution is asked to decide. Independent counsel produces the right questions. Institutional memory allows those questions to be answered from what the organisation has actually learned. Neither, on its own, ensures that the decision, when it arrives, is taken in a way that honours the work behind it. That is a question of temporal discipline rather than of counsel or of memory, and it belongs to the essay that follows this one.