AI, and in particular Agentic AI, is the latest shiny object to be applied to the goal of network automation. At the same time, telecom is on a path toward autonomous networking. It is critical to understand that these tasks are not independent; rather that agentic AI is a promising technological advance that can—and in fact must—plug into existing autonomous network control loops.

Appledore has been consistent in our portrayal of proper network autonomy for a decade. As far back as 2017 we proposed a control loop, based on intent (although we did not call it “intent” at the time), with rapid automated service assurance (“RASA”) which we also indicated would one day contain cognition. In other words, AI.

In this document we review control theory in the abstract; review its adaptation to telecom networks; look at what AI agents for telecom should look like; and illustrate some common ways that AI may be inserted into a telecom network autonomous closed loop.

This research note should be read in the context of complementary research which may be found here, describing our view of how AI agents can best exploit their most crucial consumable: data.