Jane Bennett talks quite precisely about actants in a federation. It contrasts with the image of a hierarchy of command and obedience where there is a final layer of items capable of nothing beyond following orders (normally laws). As a consequence, there is no emperor of action capable to overcome in its power all and each of the actants that contribute to the outcome. No decision is sovereign like that - it has to negotiate with whatever is available to implement it. Bennett then refers to the surprise in outcome associated to each action, as remarked by Latour. There is always a messy interplay between the human and the non-human, otherwise there is no action. Intention is never more than one actant among many and it can itself be cracked into many.
Of course the usual reply is that some items in the world simply obey laws. Often meaning that (non-human) actants are somehow lawlike while (human) agents are autonomous. This can be said because we choose some margins of error excluding the scope where the actants can go astray. The laws are taken to have this margin of error and therefore when electricity spreads in a clinamen, say, we can still present its overall trajectory as lawlike. A similar approach could be taken with human agents: their overall trajectories (considering a sufficiently large group of agents in a long period of time) is quite lawlike, it is only when we look at the small events, when we focus on the small picture, that we find the clinamens. Maybe it is an issue of focus, of which view is taken: the grand view is lawlike, swerves appear in the small picture, in the details. Usually, agents are viewed in a close-up while actants are viewed from afar in medium or long shot.
But what worries me about Bennett's mixed approach (actant theory, that is process philosophy, associated with vital materialism) is that when it comes to ontogenesis, matter can be seen itself as an actant. Harman notices that process philosophy is not itself enough to fully exorcise materialism - as he thinks one should do. Maybe. But it is a strange form of materialism this one championed by Bennett. If action is always a product of an unpredictable federation of actants, matter is either one among several others and therefore no special role can be ascribed to it or its role is fully dispensible. It seems to me that if one appeal both to vital matter and to a critical mass of actants to explain (animate) action, one offers an explanation too many. It seems then that matter itself is not the source of animation (or of vitality).