GOAP
A planner that assembles a chain of actions at runtime to reach a goal, instead of following a pre-authored behavior.
The technique (Goal-Oriented Action Planning) is a game adaptation of STRIPS planning from classical AI. Each action carries preconditions, effects and a cost; the agent holds a goal, and a backward A* search through the space of actions finds the cheapest chain that satisfies it. Unlike a state machine, where every transition is written by hand, the actions are stitched together at runtime.
Jeff Orkin popularised GOAP with the AI in F.E.A.R. (2005), whose enemies took cover, flanked and fell back without any of it being explicitly scripted. Later titles such as S.T.A.L.K.E.R. and Shadow of Mordor used similar schemes. The cost is debugging: when an agent does something surprising you have to read the plan backwards, not just inspect a single state.