Regular readers know that I embrace a simple three-perspective approach to enterprise architecture: process, data, and system. All three of these domains are seeing significant trends towards increased agility, at a fundamental level that I think is more than hype.
In the world of process, the most important trend is towards processes that may not be repeatable; that is, each instance of a process may vary dynamically - there cannot be One Process Model to Rule Them All. Towards this end, the OMG has issued an RFI to determine "whether it is appropriate to develop standards at this time for...these dynamically developed processes." This is related to but not the same as some of the debates I've been in recently regarding so-called "value networks" - in my view, much greater precision can be expected if the OMG is involved. Unanswered questions here include the data consequences of variable processes: how do we report? How do we manage by metrics when the process is so variable that metrics are difficult if not impossible to establish?
In the world of data, the most important agility trend is the Semantic Web, which can be seen as a critique of rigid, monolithic, totalizing data architecture (one Data Model to Rule Them All...) Its "open universe" assumptions and fundamental recognition of the ongoing need to reconcile multiple ontologies are the essence of agility.
I was a Semantic Web skeptic for some time, and am still concerned about the talent management implications, but we are starting to see the necessary ecosystem emerge. As far as credibility, I refer you to the pages of Scientific American, where it has had favorable coverage in 2001 and 2007. It's not a vendor-driven fad. What really convinced me was Dave Hay's favorable assessment (Dave being a notable data curmudgeon, not easily impressed by fads), and Wilshire Conferences adding a new conference dedicated to the topic.
In the world of systems, the most important agility trend is virtual appliances, as seen in CohesiveFT's products. The Seybold Group has a good analysis of this trend's importance. Any number of caveats apply, but the essential proposition is that the perpetual desire of IT hosting groups to minimize and standardize the OS and middleware stacks underpinning applications is misguided. Instead of One Technology Stack to Rule Them All, virtualize applications into finely tuned, purpose built combinations of OS + middleware + application code, and stop worrying about consistency. (Security, licensing, and talent management are some unanswered questions here...)
All of these trends have a distinct postmodern feel to them, especially Semantic Web, but the other two as well. Truth and meaning, and in particular the utility we derive from them, are context dependent. There is no grand, totalizing "theory of everything," no universal discourse that can handle all perspectives. Perhaps it is no coincidence that the Communications of the ACM recently published an article on critical theory applied to IS. All of these trends point in the direction of greater diversity and greater specialization, and away from monocultures: operational, semantic, and technical.
The core challenge: will we have the trained human capital able to wring value from this diversity? Integration specialists, those who can wrap their heads around the multiple perspectives and align them as needed, will be in high demand - but such specialists (among whom I would count enterprise architects) need to temper their need for unified frameworks, recognizing that such pursuits are often at best academic and at worst futile.