[Opinion] Architecture as expression of values
For quite sometime, I've been talking about how architecture is the expression of our values and one key value we need to discuss is where do we value human judgement in the decision making process that is software.
Obviously, I use a map (see image) to explain this, and though I used to call it conversational programming early on, these days I used the term "vibe" to keep up with the current rage. The term changed, not the concept.
The problem I face is that most organisations have no maps, they don't even understand their value chains (or the supply chains if we are talking across an industry). Without contextual understanding, this leaves plenty of room for the magic one size fits all method, the "SAFe of AI".
For reference, I was invited as the "chaos monkey" to the SAFe leadership summit many years ago. I pointed out that SAFe would have a context where it worked but it wouldn't work everywhere (see XP, see Six Sigma etc). Naturally, this fell on deaf ears and as usual we got the rounds of people blaming others for not using SAFe components correctly rather that identifying the inherent weakness in the system itself.
For example, my maps have a very limited context, they are useful in understanding a landscape but they don't tell you what to do or how you should do it. Those actually require thought applied to the map.
I find that understanding context is useful because without that most organisations tend to be seduced by one size fits all. It's almost inevitable in this AI world.
For reasons of clarity, by "AI", I'm referring to LLM/GPT as that has sucked most of the oxygen out of the room. Yes, I'm aware AI is a field with many components, many of which are highly useful in specific contexts. I mapped that long ago, as have many others. If you want a map on AI, I'd probably go talk to Mark C. ... though be warned, he might have used an LLM/GPT to generate it. That's not a bad thing, but just review and challenge it (as with all maps). Remember, all maps are imperfect representations of a space, in fact they have to be in order to be useful.
Anyway, whilst we are going through the process of developing practices in this space, I'd be grateful if you could point out any "one size fits all" methods of AI that you see. That could be books or claims of "this works everywhere", "works for every organisation" and "no need for architects anymore". My interest is not that they are wrong (as one size fits all, they will be) but in what contexts they might apply.
Originally published on LinkedIn.