You might be interested in Tomasello's "The Evolution of Agency" where he kind of addresses this very question. It really depends on how you define "making decisions" and "purely chemical reactions" doesn't it - all life is chemical reactions, including when we make decisions, and it's easy for us to apply decision-making language even to systems that are simple enough that we can see them as "purely chemical reactions".

Tomasello defines the notion of "agents" as "feedback-control systems" that he distinguishes from pure stimulus-response systems. In his examples a nematode for example is "stimulus-response"; its behavior is very directly related to its immediate environment. If it runs into food it eats, otherwise it doesn't, and there isn't really a notion of it seeking out food when it's hungry and not when it's not. In contrast and "agent" is a feedback-control system with goals, a perceptual system that checks whether the goal is accomplished at any given time and a behavioral repertoire aimed at accomplishing the goal. In our lineage he sets the appearance of this agency around the evolution of vertebrates, and uses lizards as an example of the most basic level. (he doesn't address other lineages other than to say that various levels of agency clearly evolved convergently a few times; so octopuses and social insects for example would also have these systems). So where a nematode has feeding behavior that's triggered by running into food and other behaviors when food isn't present, a lizard's behavior depends not only on the immediate stimulus but on more abstract goals - in a given environment it might be currently hungry and looking for food, or sated and looking for shade or sun to rest or hide or thermoregulate, or looking to reproduce, etc, and its behavior will depend on and be directed towards accomplishing that goal.

It's interesting that you say "thinks through and makes decisions" as if they're on the same level but the book actually claims that human agency is actually the result of the evolution of several successive layers of feedback-control mechanisms that each allow more flexibility and responsiveness - so for example lizards have a feedback systems that adjusts behavior to achieve goals, and mammals have that and also a higher-level feedback system above that to adjust the goal-seeking behavior itself, mentally "playing out" different ways of accomplishing the goal in order to pick the best one. He describes four such levels for humans and it suggests a variety of ways we could define "think through and make decisions", with different species qualifying or not depending on which we choose.