The OODA loop was developed by military strategist John Boyd in the 1970s to describe how fighter pilots — and any skilled decision-maker — stay effective under fast-moving, unpredictable conditions. The four stages are: Observe (take in new information from the environment), Orient (make sense of it, filtered through everything you already know), Decide (choose a course of action), and Act (execute it). The loop then repeats. The crucial insight is that the loop itself is the unit of performance, not any individual decision. Whoever cycles through it faster than their situation changes, stays ahead.
AI agents run a version of this loop continuously — and it is the mechanism that separates them from simpler automated systems. A script or workflow observes once at the start and executes a fixed sequence regardless of what changes. An agent re-enters the loop at every step: it observes the current state of the task, orients against its goals and constraints, decides on the next action, executes, and then immediately observes again. When something unexpected happens — an item is out of stock, a file is missing, new information arrives — the agent catches it in the Observe stage and adjusts before the situation can outpace it. The loop is what gives agents their ability to adapt rather than just execute.
Of the four stages, Orient is the most consequential and the most underappreciated. Boyd called it the centre of gravity of the whole loop — the place where prior knowledge, mental models, assumptions, and context all bear on new information at once. For a human decision-maker, Orient is shaped by experience, training, and judgment built over years. For an AI agent, Orient is shaped by its instructions, its memory, and everything it has been given as context. This is why the quality of an agent's instructions matters so profoundly: you are not just telling it what to do, you are shaping how it makes sense of what it sees. An agent with excellent execution and a poorly calibrated Orient stage will do exactly the wrong thing with great confidence and speed.
The loop is worth understanding not just because AI agents use it, but because it describes something true about intelligence under pressure more generally. In a world where agents handle execution at scale, the distinctly human contribution shifts toward the Orient stage: providing the context, values, and judgment that shapes how the agent makes sense of what it observes. The people who use AI most effectively are not necessarily the ones who act fastest. They are the ones who orient most clearly — and who understand where in the loop their role actually lives.
