I think there's a lot to this line of thinking. It's in fact the counterargument I find most threatening to my position.

But I think you are assuming an organization with a particularly autocratic leadership. In some organizations, leadership is broadly distributed.

For example, in many open source software development communities, decisions about how to change the source code are made by a consensus of their developers.

When these developers are using their own software in the process of developing and/or communicating (such as in the case of Git, or Mailman, or Emacs), then I think there's a case for a genuine, distributed sense of organizational intelligence with recursive self-modification.