Knowledge management never actually replaces some types of knowledge.
When I say knowledge management, I mean any effort to record, organize, and store knowledge related to some project. Source code is part of this record. And while such records are certainly very important to keep, they do not guarantee the a project can be done again from scratch with the same result, whether successfully or otherwise.
The lost art
A while ago Pentax, Japanese camera maker, announced they would start making film cameras again. In the announcement video for their project they explained why they were doing this:
In years since Pentax made the first film cameras, many engineers knowledgeable about film cameras have retired. Pentax [...] has always kept complete collections of blueprints, documentation, and similar material from that period. However, you cannot make a film camera just with that. There is expertise in the production methods and many techniques and tricks that cannot be written into the blueprints, that have been verbally passed on by veteran engineers through instruction, generation by generation. Without these, a film camera cannot be made. [...] This is our last chance [...] In a little while, all engineers who know film cameras will be [retired] and I think Pentax will not be able to make a film camera ever again."
(Takeo Suzuki, product designer, Pentax)
This is what we commonly recognize as tacit, or implicit knowledge, things that can only be learned through direct experience in training or independent exploration. A big part of why expert practitioners are able to operate at very high levels of performance is precisely because they don't necessarily take time to verbalise every decision they make. A lof it is therefore never recorded, and sometimes cannot even be recalled, even though it clearly works.
Combating the loss of expertise
Amids the waves of layoffs, people are thinking about ways to reduce the damage of experts leaving the force. As usual, recording their knowledge sounds like a very obvious way to mitigate the damage. Unfortunately, reducing the impact of layoffs by doing systematic knowledge management will ultimately result in a huge database of useless documents in the absence of proper experience transfer.
When there are no veteran engineers who can train the staff, the history of knowledge is interrupted. The remaining staff has no other choice by to retrace the steps of their predecessors on their own and from scratch, a long and arduous process with no guarantees of ever reaching the same results as before.
How to prevent the loss of expertise
The only way loss of expertise can be prevented is to replicate the tacit knowledge. This can only be done between humans that share the same context (work on the same project, for example), and intermediaries like written document do not suffice. Even presentations are not sufficient.
What actually needs to happen is to get two people to share the context as much as possible, and then have them observe each other while working. Pair coding is a common practice to facilitate the transfer of implicit knowledge, but it can also be done in the form of explicit training.
Because this process takes time and a positive attitude, when the layoff is already happening, it is way too late to think about it. It has to be done on an ongoing basis right from the first day. It is never too late to start, though.
