But maybe that’s the wrong question. Maybe a better question would be this: Can a change-dominated world even exist without some kind of built-in change management?
Change management is always an attempt to impose orderly processes on disorder. That, at least, doesn’t change. What does change is the nature and the scope of the disorder, and the nature and the scope of the processes that must be imposed on it. This is what makes the DevOps world look so different, and appear to be so alien to any kind of recognizable change management.
Traditional change management, after all, seems inseparable from waterfall and other traditional development methodologies. You determine which changes will be part of a project, you schedule them, and there they are on a Gantt chart, each one following its predecessor in proper order. Your job is as much to keep out ad-hoc chaos as it is to manage the changes in the project.
And in many ways, Agile change management is a more fluid and responsive version of traditional change management, scaled down from project-level to iteration-level, with a shifting stack of priorities replacing the Gantt chart. Change management’s role is to determine if and when there is a reason why a task should move higher or lower in the priority stack, but not to freeze priorities (as would have happened in the initial stages of a waterfall project). Agile change management is priority management as much as it is change management — but it still serves as a barrier against the disorder of ad-hoc decision-making.
In Agile, the actual processes involved in managing changes and priorities are still in human hands and are based on human decisions. DevOps moves many of those management processes out of human hands and places them under automated control. Is it still possible to manage changes or even maintain control over priorities in an environment where much of the on-the-ground decision-making is automated?
Consider what automation actually is in DevOps — it’s the transfer of human management policies, decision-making, and functional processes to an automatically operating computer-based system. You move the responsibilities that can be implemented in an algorithm over to the automated system, leaving the DevOps team free to deal with the items that need actual, hands-on human attention.
This immediately suggests what naturally tends to happen with change management in DevOps. It splits into two forks, each of which is important to the overall DevOps effort. One fork consists of change management as implemented in the automates continuous release system, while the other fork consists of human-directed change management of the somewhat more traditional kind. Each of these requires first-rate change management expertise on an ongoing basis.
It isn’t hard to see why an automated continuous release system that incorporates change management features would require the involvement of human change management experts during its initial design and implementation phases. Since the release system is supposed to incorporate human expertise, it naturally needs expert input at some point during its design. Input from experienced change managers (particularly those with a good understanding of the system being developed) can be extremely important during the early design phases of an automated continuous release system; you are in effect building their knowledge into the structure of the system.
But DevOps continuous release is by its very nature likely to be a continually changing process itself, which means that the automation software that directs it is going to be in a continual state of change. This continual flux will include the expertise that is embodied in the system, which means that its frequent revision and redesign will require input from human change management experts.
And not all management duties can be automated. After human managers have been relieved of all of the responsibilities that can be automated, they are left with the ones that for one reason or another do not lend themselves well to automation — in essence, anything that can’t be easily turned into an algorithm. This is likely to include at least some (and possibly many) of the kinds of decision that fall under the heading of change management. These unautomated responsibilities will require someone (or several people) to take the role of change manager.
And DevOps change management generally does not take its cue from waterfall in the first place. It is more likely to be a lineal descendant of Agile change management, with its emphasis on managing a flexible stack of priorities during the course of an iteration, and not a static list of requirements that must be included in the project. This kind of priority-balancing requires more human involvement than does waterfall’s static list, which means that Agile-style change management is likely to result in a greater degree of unautomated change management than one would find with waterfall.
This shouldn’t be surprising. As the more repetitive, time-consuming, and generally uninteresting tasks in any system are automated, it leaves greater time for complex and demanding tasks involving analysis and decision-making. This in turn make it easier to implement methodologies which might not be practical in a less automated environment. In other words, human-based change management will now focus on managing shifting priorities and stakeholder demands, not because it has to, but because it can.
So what place does change management have in a change-dominated world? It transforms itself from being a relatively static discipline imposed on an inherently slow process (waterfall development) to an intrinsic (and dynamic) part of the change-driven environment itself. DevOps change management manages change from within the machinery of the system itself, while at the same time allowing greater latitude for human guidance of the flow of change in response to the shifting requirements imposed by that change-driven environment. To manage change in a change-dominated world, one becomes the change.
Complete visibility for DevSecOps
Reduce downtime and move from reactive to proactive monitoring.