KM Europe 2003
Amsterdam RAI Centre 10-12th November

I attended the first 2 days of this event, which they claim is the world’s largest. The format is a free trade show and a number of free talks, and then a fee per keynote presentation. It was well attended (they had 1500 last year and probably about the same this time) though the vendor booths often looked very quiet. Here are some observations on things that stood-out:

Vendors
None of the products on display really looked like anything new – still the same old search and data management tools on the whole.
Entopia, with its bottom-up approach to KM seems to be taking off. I was disappointed with their Social Network Analysis offering, as its purely about visualising document sharing (probably the least of all possible networks you can map). They didn’t seem to have any proper network analysis tools behind it either. The only good thing is that its updated automatically, unlike the resource-intensive ‘snapshot’ produced by questionnaire methods.
AnswerWeb is a relatively new Dutch player in the Expert Location\Q&A sector (along with Askme and Sopheon – formerly Organik). It seems to lack the automatic profiling of people’s expertise though.

Presentations
(these are now online at: http://www.kmeurope.com/presentations.asp )
KM at Renault: Jean-Marc David
This is the first time I've heard about KM from Renault, but they actually have a very mature programme with roots in AI and Expert Networks in the 80's. Their current programme is “one of Renault's top 100 strategic initiatives” [top 100?!], based in a Business Transformation branch of their central IS/IT function.
Their approach reminds me of Ford, Rolls-Royce or Siemens: a manufacturing R&D best practice database with a mandate that employees contribute. Technical Domain Leaders do quality reviews before publication.

 Renault's suggestion scheme has 350, 000 contributions/year, saving €57M in 2002! This seems very much down to their culture of individual initiative and creativity to try things out.
 Collaboration: they felt it necessary to build an in-house tool to re-use existing systems e.g. Documentum, task management, issue management, news etc. rather than make the case for a whole new suite. In the end they co-built it with Nissan for both internet and Extranet use.
 Renault tried providing different template virtual collaboration spaces to meet the different needs of Communities of Interest, communities of practice, “micro-orgs” and project teams. But they found it didn't work as users were not comfortable with these distinctions, so now they take a more bespoke route of working with each client to define best mix of functionality for them.

David Snowden - Complex Knowledge - IBM Cynefin Centre
Snowden was on fine form, as always:
“We now know enough about group dynamics that we can compress the effects of 3-4 years ad-hoc network formation into 3-4 months of deliberate network creation. [i.e. its not that people are bad at networking, but that they're very often sub-optimal and now we know how to do something about it in a way we couldn't 10-15 years ago]. This is not the same as taking an informal network and making it formal - a mistake CoPs sometimes make, and in doing so kill of what was working already. e.g. one way is to find mavericks and let them self-organise a community.”

On the “US vs. Europe/Asian approach”. The US tries to find practice leaders and replicate wholesale. Europeans try to look at good and bad cases and find common principles. People learn very readily from worst practice because on the whole avoiding failure is a better survival strategy than only focussing on success.

This leads to:
* Rules: try to work out all possible events and define response
vs.
* Heuristics: these tolerate ambiguity but are less clear about when they apply exactly
We do need both, depending on context. [cf. laws vs. ethics]

KM has too often tried to manage by rules: design a system around an ideal set of behaviours and then put in place a change-management system to make people behave in the ideal way.The alternative, necessary for any complex system according to Snowden, is to use 'Boundaries' and 'Attractors'. With kids you have firm rules and come down heavily if crossed, but mostly try to keep kids away from the boundary by attractors such as football and food. When undesired patterns begin to emerge (e.g around vodka) you step in an disrupt the pattern.
“So why do organisations try to manage their employees in the 'rule' sense? When faced with a new task, would you rather go to the Corp. Best Practice database and follow a document, or talk to 4-5 people who have done it before and find out what happened? So why build 'knowledge bases'?” [well, one reason is churn – sometimes its hard to find anyone still available to talk to, let alone 5. Narrative databases may be the right compromise though - Sam].

Social Complexity:“Human Beings are not Ants!” Ants are condemned to always act the same (complex, but 'rational') way. Humans have free will and can choose to behave in a non-complex way by e.g. creating and following structures.

Contingent Complexity:
Snowden, having mocked consultants for their 2x2 matrices, then produced on of his own:
*Visible Order: cause and effect obvious to all (sense, categorise, respond = best practice). OK to manage this by a formal structure because the system is visible to all.

*Hidden Order: Cause & Effect is discoverable by experts (Sense, analyse, respond = good practice). Manage by tightly connected peers as well as central control.

*Complex Un-Order: Cause & Effect is coherent in retrospect (Probe, sense respond). The risk is that this looks like Hidden Order but its not, the C&E is only visible once the pattern emerged. There are just too many possible connections to predict analytically until its over. 9/11 investigations are falling into the trap of thinking that if they just throw enough analytical power in, then they will be able to detect the next patter in advance – they won’t. Manage by tightly-connected peers and loose central control. Worst-practice sharing works well here.

*Chaotic Un-Order: no perceivable Cause & Effect even after event (act quickly, then sense and respond).

In crisis management (Chaotic un-order), usually leaders step in and create a network around themselves to stabilise things (Visible order). But this is very brittle, so its better to move into Complex Un-Order by rapidly creating a peer network, then look for patterns you want to reinforce and ones to suppress as this leads to much more robust order.

IBM found that if employees can create their own informal communities, then the resulting number is roughly half number of staff! There is no way you could formally intervene to design all these, you do it by using attractors so that others 'swarm' round them.

Knowledge Acquisition & Modelling Process (KAMP) - Rolls-Royce
Michael Moss gave an overview of R-R's excellent programme for capturing engineering expertise. They do this by training and coaching graduates in the toolbox of techniques that rose out of Expert Systems (ES) in the 80's/90's. These are effective but little-known in the KM World. They fell into disuse because it was very costly to produce an ES, but if you stop at the documentation stage, its viable. Even better, its time-effective as graduates learn more quickly and experts typically only have to give up about 12 hours over 3 months. They gain by making answers to routine questions readily available, and the organisation gains a lasting resource that gives some protection against loss through retirement or moves.

What really makes KAMP work though is not just the capture bit, but the whole context of doing reviews to identify vulnerable knowledge, securing an owner for the output, having an established dissemination route (the ‘Capability Intranet’ they call it) and an ongoing quality/maintenance procedure. Over 130 have been done so far.

Verna Allee - Knowledge, Networks & Value-Creation
"We're learning that to market its much more effective to target hubs than to try to hit a whole network equally"

"People think Silicon Valley is really innovative but it isn't - which is why few survived the dotcom crash. They are good at technology innovation, but not at social innovation to adapt their businesses (e.g. moving from in-house R&D to network collaboration is what Verna means by social innovation). Nor do they innovate Business Analytics e.g. managing dynamic, intangible systems - they're still trapped in static business models. The survivors like Amazon and ebay are still around because they were innovating on all 3 fronts."