The budget System 2 runs on
Deliberate reasoning, System 2 in the usual shorthand, is not free. Holding several possibilities in mind at once, updating a plan against new information, and noticing your own reasoning is starting to slip all draw on the same limited resource: working memory capacity, which is metabolically expensive and highly sensitive to physiological arousal. This is not a metaphor. Cognitive load studies going back decades show working memory performance degrading measurably under acute stress, and the degradation is not random, it is specific to exactly the tasks System 2 handles: holding multiple variables, inhibiting a tempting first answer, revising a plan mid-stream.
System 1, by contrast, barely notices. Pattern recognition, gut reads, and fast heuristic judgment run on machinery that evolved to work under threat, because that is precisely when an organism cannot afford the latency of deliberation. So the failure mode under pressure is not "the executive stopped thinking." It is "the executive kept thinking, in the one mode that was never designed to catch its own mistakes."
Where the gate actually sits
The honest version of this claim needs a caveat up front: the detailed wiring of how autonomic arousal gates cognition, the specific pathways, the role of the vagus nerve, the exact timing, is an active and contested area of research. Polyvagal theory in particular, popular in wellness and coaching circles, makes stronger anatomical claims than the current evidence fully supports, and I try to be careful about which parts of that framework are well established versus which are a plausible but unproven extension. What is well established, across a much less contested literature in stress physiology and cognitive psychology, is the pattern itself: acute physiological arousal reliably narrows attention, degrades working memory, and biases judgment toward fast, previously reinforced responses. The precise mechanism is still being mapped. The effect is not in serious dispute.
Heart rate variability, the beat-to-beat variation in time between heartbeats, is one of the better-validated windows onto this state. Higher HRV correlates with better performance on tasks that specifically require System 2: inhibitory control, cognitive flexibility, emotion regulation under provocation. This is why HRV biofeedback, structured practice with slow paced breathing to increase variability, has a real evidence base in clinical and performance psychology, distinct from its much looser use as a wellness buzzword. The mechanism worth trusting is the correlational and interventional evidence: raise HRV before a demanding cognitive task, performance on that task tends to improve. The mechanism worth holding more loosely is the specific neuroanatomical story for why, which is still being worked out.
What this means at the moment of decision
Put the executive back in the room. A reorg is going wrong in real time, a board member is pushing back harder than expected, a term sheet just changed under pressure to close. The physiological state at that moment, not the executive's intelligence, not their experience, is what determines whether System 2 is actually available to be called on. Under high acute arousal, it usually is not, however capable the person is when calm.
This is where AI changes the failure mode rather than fixing it. An executive under load reaches for a tool, asks a question, and gets back an answer: confident, well structured, apparently reasoned. Because System 2 is the part of the mind that would normally interrogate that answer, and System 2 is exactly what is offline, the AI's output gets accepted at the fluency of System 1: fast, unexamined, and mistaken for having been checked. This is not a claim about AI being persuasive or manipulative. It is a claim about timing. The tool arrives at precisely the moment the executive has the least capacity to evaluate it, and it is good enough at sounding right that the absence of evaluation goes unnoticed.
The practical implication
None of this argues for suppressing System 1 or distrusting intuition generally, System 1 is right often enough that overriding it indiscriminately would be its own mistake. The practical implication is narrower and more useful: before a high-stakes call, the relevant question is not only "what does my gut say" or "have I reasoned this through." It is "what is my actual physiological state right now, and does it match the kind of thinking this decision needs." A five-minute check, informally or with an actual HRV reading, before walking into the room that matters is a cheap way to answer a question most executives never think to ask.
That is a different kind of preparation than the content of the decision itself. It does not require more information, a better model, or a smarter tool. It requires noticing which system is actually available before trusting what it tells you, and building the habit of checking rather than assuming the answer.