Wednesday, January 19, 2011

How to combine Enterprise Architecture with Systems Thinking

#entarch #systemsthinking A great discussion yesterday between some systems thinkers from SCiO and some enterprise architects. This post is not a record of the meeting, but contains some of my thoughts following the meeting.

Does it make sense to combine EA and ST to create something we might call EAST?
Let me address this question in terms of the five perspectives outlined in my previous posts.

EAST as instrument Use ST to improve EA. Use EA to improve ST. Or create a composite instrument using elements of EA and ST.
EAST as discourse
Either find the intersection between EA and ST - common concepts and principles. Or find the union between EA and ST - complementary concepts and principles.
EAST as community
Encourage EA people to become part of the ST community, or vice versa. And/or facilitate collaboration between EA people and ST people.
EAST as knowledge
Use ST to investigate and critique EA. Use EA to investigate and critique ST.
EAST as trade or service
Use EA as a platform for introducing ST into organizations. Use ST as a platform for introducing EA into organizations (perhaps at a different level).

All of these combinations are theoretically possible. It is possible to take any two or more disciplines and create an arbitrary hybrid, as long as you have pretty weak standards of coherence and usefulness. Weak coherence may be an essential step for early innovation, but shouldn't be an excuse for intellectual laziness. Gartner is currently pushing a specific hybrid, based on combining EA with design thinking together with some ecological ideas (panarchy), but I haven't seen any practical results yet. See my post on Hybrid Thinking and my note on Methodological Syncretism.

The practical question is how any such possible combinations can be grounded in practical work, rather than being merely abstract combinations of ideas and slideware. So the next task for the EAST group is to find some practical business problems to work on together.


Alexander Samarin said...

A thought how they can work together. Considering that EA does a great job in describing the “enterprise genotype” (a full nomenclature of enterprise artefacts or assets) and there are many techniques to evaluate the “enterprise phenotype” (a set of observable characteristics such as performance), then the ST with its executable models of relationships between artefacts can form the bridge (an enterprise executable model with the use of BPM and SOA) from the “enterprise genotype” to the “enterprise phenotype”.


Richard Veryard said...

Thanks Alexander. The genotype/phenotype distinction is an interesting one. A given nomenclature of enterprise artefacts or assets can only be regarded as "full" or "complete" relative to a given EA discourse. For example, followers of the Zachman doctrine are convinced of the completeness of the Zachman framework, and cannot conceive that there might be anything else worth paying attention to. (See my presentation on the Kipling Zachman lens.)

Some schools of ST (for example systems dynamics and first-order cybernetics) might give us executable models of relationships which would be complementary to the models offered by EA. Your proposal appears to regard EA models and ST models as separate tools (instruments) that can be put together to achieve a larger effect.

Other schools of ST might work directly on the EA models, questioning their completeness and fit-for-purpose, and helping to frame the conceptual relationship between genotype and phenotype. Thus we are using ST (among other things) as an instrument for refining EA-as-instrument.

There are also important questions about the observability of such characteristics as performance, and the extent to which the way we choose to measure these characteristics is influenced by our conceptual models (genotype).

But I think the big question here is the nature of the discourse in which EA and ST can work together in the way you describe. To what extent does the EA genotype determine the ST phenotype, and what other phenomena might emerge from looking at the enterprise from multiple ST perspectives.

Alexander Samarin said...

Thanks Richard.

I like “extended interpretations” in your presentation.

Concerning executable models of relationships: Maybe business processes are good candidatures for such models – each business process is a relationship between events, data, documents, roles, rules, services, business process, audit trails, KPIs, etc. BPM considers together modeling (or planning), automation (or instrumentation), execution, control, measurement and optimization of business processes. So, a process-managed enterprise has its executable model “by definition”.


enectoux said...

Hello Richard,

To me EA and ST cannot be dissociate. Thinking about my own experience, I started to use ST way before I came to EA, but now I don't see how we could separate them. Just to take a silly example: practicing EA without having ST approach, you will necessary fail. How is it possible to do EA without considering surrounding contexts and interactions of the different component.

So, I don't see the need to combine them, mixing EA &. ST, because they are from the begining.

Richard Veryard said...

Thanks Emeric

I'm not asking for EA and ST to be separated, or even for their combination to be justified, merely for us to think about the many different ways in which they can be combined. There are of course many different flavours of EA and many schools of ST, which produces a large number of possible permutations.

Of course I should hope that most enterprise architects would be capable of "considering surrounding contexts and interactions of the different component", but there is a lot more to systems thinking than that.

Peter said...

I believe that ICT/Enterprise architects will be forced to use some kind of ST due to the increasing complexity. Static diagramming tools like ArchiMate won't be sufficient to understand complexity. Simulation tools like system dynamics and/or agent-based modeling shall be necessary. I (as an infrastructure architect) am looking at this moment at the possibilities of NetLogo which can do both... (although the system dynamics modeler is a bit rough)

Ivo said...

Peter, InsightMaker can do both as well, but there the agent-based simulation is still very basic. And then there is Anylogic which seems not only good at both offering process simulation as well. Unfortunately, unlike Netlogo and Insightmaker(IM), Anylogic is not free. I personally use Vensim for SD, sometimes IM, and ARIS for process simulation integrated with other EA models. My interest to NetLogo is growing thanks to Melanie's course.