The Warehouse Was Never the Problem — Why Knowledge Management Captured Everything Except the Knowledge
We took a theory whose whole insight was that the valuable knowledge could not be written down, and we implemented the one-twentieth of it that a piece of software could execute.
The repository that filled up and emptied out
Somewhere on the corporate intranet, two clicks below the home page, there is a knowledge base. It was built about two years ago with real budget and real conviction — a searchable repository into which every project team was to deposit its lessons, every specialist their methods, every expert the judgement that until then had lived only in their heads. There was a launch event. There was a mandate from the top. For a quarter or two the numbers climbed handsomely: four hundred documents, then a thousand, then the better part of four thousand. The programme reported green throughout.
Now ask someone who actually needs to know something today — how we handled the migration on that difficult account last year, who really understands the settlement interface, what went wrong the last time we attempted this — where they go to find out. They will not say the knowledge base. They will say they asked someone. They walked over, or picked up the phone, or remembered that a particular colleague had been through something similar and went to find them. The repository, meanwhile, holds four thousand documents and answers almost none of the questions anyone actually brings to it.
This is not the story of one failed system. It is very close to the universal experience of the knowledge management movement that has swept through large organisations over the past five years, and it is worth being honest about now, in the sober light of the spring just past, when a great deal of confident spending is finally being re-examined. We were sold, and we bought, the idea that knowledge was a corporate asset that could be captured, stored and retrieved like any other — and that with the right platform we would at last stop losing it. The platforms were duly installed. The knowledge kept walking out of the door exactly as before.
The knowledge management programme of the late 1990s solved a logistics problem almost nobody had, and left untouched the human problem almost everybody had.
We mistook the map for the territory
The intellectual case for knowledge management was not foolish. Anyone who read Nonaka and Takeuchi understood, at least on the page, that the crucial distinction was between explicit knowledge — the kind you can write down — and tacit knowledge, the kind that lives in judgement, instinct and relationship and resists being written down at all. Davenport and Prusak had said it plainly enough. The entire point of the theory was that the hard part was the tacit part.
And then, in practice, we spent our money almost entirely on the explicit part. Of course we did. Explicit knowledge is the part a system can hold. So a subtle argument about how human beings come to know things arrived in the building as a server, a document taxonomy, a mandatory “lessons learned” template bolted to the end of every project and — in the more ambitious organisations — a Chief Knowledge Officer with a mandate and a launch budget. We took a theory whose whole insight was that the valuable knowledge could not be written down, and we implemented the one-twentieth of it that a piece of software could execute.
The result is the repository described above: technically a success, practically inert. It is full of precisely the knowledge that was least worth capturing — the sanitised end-of-project report written defensively so that nothing in it could later be used against its author; the method document that lists the steps but not the fifty judgements that decide whether the steps work; the “best practice” that was true in one context and is quietly wrong in yours. What it does not contain, because it structurally cannot, is the thing you actually needed: the reason it went the way it went, which lives in a person, and which that person will give you in ten minutes on the telephone and would never, under any circumstances, type into a form.
“A repository holds what people already know they know. It is helpless before the knowledge they only produce when someone asks them the right question.”
Why the substitution was so easy to make
If the failure is this predictable, why did so many capable organisations walk into it together? Not through stupidity. The pull towards the technological version of knowledge management, and away from the human one, was structural, and every strand of it was individually reasonable.
- You can buy a platform; you cannot buy a culture. A vendor can sell you a knowledge base and have it installed by the end of the quarter. Nobody can sell you an organisation in which experts are willing to share what they know. Faced with a problem framed as “we keep losing knowledge,” the buyable answer will always beat the unbuyable one, because it can be procured, scheduled and shown to the board.
- Capture is auditable; trust is not. You can report that four thousand documents were contributed and that ninety per cent of projects completed their lessons-learned template. You cannot report that two engineers who did not previously trust each other now do. Every metric that governance understands is a metric of deposit, not of use — and so the programme optimises, quietly and relentlessly, for filling the repository rather than for anyone getting anything out of it.
- The technology was the fashion. This has been the decade of the enterprise system. We have just come through the largest coordinated technology effort in corporate history to survive the date change, on the back of years of enterprise resource planning rollouts that taught a generation of managers that the answer to any organisational problem is an integrated platform. Knowledge management arrived dressed in those same clothes, and we recognised the outfit and reached for our wallets.
- Codification felt like progress; conversation felt like waste. An expert typing up a document produces a visible artefact. Two people talking at the coffee machine produce nothing you can point to, and in a culture that measures utilisation it looks like slack. So we industrialised the part that leaves a trace and left the part that actually moves knowledge to happen on its own — unfunded, unmeasured and faintly disapproved of.
None of these forces is malign. Together they produced a discipline superbly equipped to store the knowledge nobody needed and structurally blind to the knowledge everybody did.
What actually moves knowledge
It is easy to be clever after the fact about a failed system. The harder and more useful question is what the exceptions had that the repositories did not — because knowledge does move through organisations, reliably and every day, just not through the channels we funded.
Consider the exchange the knowledge base was meant to replace. A project manager, stuck, needs to know how a particular kind of client tends to behave when a programme starts to slip. She does not search the repository; she thinks of the one person who has lived through it three times, and she goes to him. He tells her — not the sanitised version but the real one, including the political detail he would never commit to writing. Why does he tell her? Because she has helped him before, or will; because it costs him little and raises his standing; because she asked him a specific question he found genuinely interesting; because they are close enough, in the building and in trust, for the exchange to happen at all.
Every element there is exactly what the repository lacks. Look at what the transfer actually required:
- A relationship. The knowledge moved along a line of trust that already existed. Strangers do not hand one another their hardest-won judgement, and a database is a room full of strangers.
- A specific question. Tacit knowledge does not sit in retrievable units inside someone’s head waiting to be uploaded. It is produced, on demand, in response to a particular problem. The expert did not know that he knew the answer until he was asked.
- A reason to share. Reciprocity, reputation, plain goodwill — some live incentive made telling worth his while. The lessons-learned template offered him none of these, only exposure.
- Proximity. He was findable. She knew he existed, knew roughly what he knew, and could reach him. In an organisation where people cannot see one another, the most sophisticated repository ever built does not help, because you cannot put a question to an expert you do not know is there.
Not one of these is a technology problem, and not one is solved by a larger database. They are problems of trust, incentive, proximity and the plain social permission to spend time talking. The knowledge management that works is the deliberate cultivation of those conditions — what a handful of the more thoughtful writers have lately begun calling communities of practice: groups of people who do similar work and are given the time, the standing and the connective tissue to learn from one another. That is a management and cultural intervention wearing no technology at all, and it is exactly the part we declined to fund, because it could not be bought, installed or counted.
The strongest case for the thing I am criticising
There is a serious objection to all of this, and it deserves its strongest form rather than a straw one.
It runs like this. Codification manifestly does work. The large professional-services firms — the consultancies and accountancies whose whole business is selling knowledge — have built formidable competitive advantage on exactly the kind of repository I am disparaging: reusable methods, proposal libraries, engagement templates, a junior consultant in one country standing on the documented shoulders of a hundred prior engagements. If capture were as futile as I claim, those firms would not have spent the past decade and a fortune building it, nor would it be paying them back so handsomely. Explicit knowledge scales; a conversation does not. You cannot telephone a colleague for every question in an organisation of forty thousand people. At some point knowledge has to leave the heads of individuals and become something the institution owns, or it dies with every resignation and every reorganisation.
This is right, and it matters that it is right. Where knowledge is genuinely codifiable — stable, repeated, reducible to a method that survives being written down — codification is not merely possible but essential, and the firms that did it well have the results to prove it. The error was never in codifying. The error is twofold, and worth stating precisely.
First, even the successful repositories succeeded because of a culture, not instead of one. In the firms that made codification pay, contributing to and drawing from the knowledge base was woven into how people were promoted, how engagements were staffed, how professional standing was earned. The technology worked because the human system around it made maintaining it worthwhile. Strip that culture away and you are left with what most of the rest of us built: the same software, empty of trust and therefore empty of use. The tool never created the culture. The culture was the achievement, and the tool was merely its instrument.
Second, we applied the codification model universally — to knowledge that was contextual, fast-moving and political, the very knowledge that loses its meaning the instant it is detached from the situation that produced it. A stable methodology can be shelved. The reason a particular transformation stalled last year cannot; by the time it has been written down, sanitised and cleared for the record, the only true and useful parts have been edited out. We took a model that works for the codifiable minority and imposed it on the tacit majority, and then blamed people’s “reluctance to share” when it failed.
The question was never whether to capture knowledge. It was which knowledge will hold still long enough to be worth capturing — and the honest answer is: far less of it than we bought servers for.
What the reckoning makes possible
There is, unexpectedly, an opportunity in the present moment. The money has changed character this year. The easy capital that funded a hundred confident platforms has thinned dramatically over the past few months, and every large investment is being asked, belatedly, what it actually returned. Knowledge management as we have practised it will not survive that scrutiny, and it should not.
But the discipline can survive the loss of its platform, if we will let the platform go. The underlying task is entirely real: organisations do lose knowledge, do repeat their mistakes, do let hard-won expertise walk out of the door and then pay to relearn it. What we have discovered, at considerable expense, is that this was never a storage problem. It is a problem of whether the people who know things have relationships with the people who need them, reasons to help, and the plain permission to spend an hour doing so.
That is unglamorous work. It funds no launch event and impresses no board with its architecture. It looks like putting the right people within reach of one another, protecting the time in which they talk, rewarding the expert who develops others rather than only the one who bills the most hours, and treating the conversation by the coffee machine not as slack to be squeezed out but as the most productive knowledge-transfer mechanism the organisation owns. When someone next proposes that we solve our knowledge problem by buying somewhere to put it, the right answer is the one these past five years have bought and paid for: we have tried filling the warehouse, and the warehouse was never the problem. We built it for a kind of knowledge that only ever lived in people, and only ever moved between them when they had reason to trust one another. No quantity of storage was ever going to change that. It is time we spent on the reason instead.