In 2025, OpenAI granted employees stock packages worth an average of 1.5 billion Korean won — roughly $1.1 million — each, an unprecedented figure for a company that hasn't gone public. That same year, several of OpenAI's senior staff still left, accepting offers from Meta and Anthropic.
Past a certain point, compensation loses its power to keep people. The bigger problem is continuing to raise pay without understanding what actually works once pay stops working.
The Moment Someone Decides to Quit Is Almost Always a Single Event
Studies examining the gap between the intent to quit and the act of actually quitting turn up a recurring pattern. The moment someone decides to leave a company is rarely the point where accumulated frustration crosses some threshold — it's usually triggered by a specific incident. Researchers call this a "shock event."
In their public statements, senior staff who left OpenAI rarely cite salary gaps or the size of their options as the main reason. Phrases like "I don't agree with the direction" or "the way I could contribute had changed" come up again and again. There's no way to verify from the outside whether these are genuine reasons or just polite cover. But the fact that departures kept happening even at the industry's highest pay levels makes it hard to dismiss these statements as mere rhetoric.
Shock events tend to fall into recurring types: having your contribution credited to someone else, watching your decision-making authority shrink without warning, or sensing that the organization's direction has drifted from the reason you joined in the first place. All three show up regardless of company size.
All a $1.1 Million Stock Grant Does Is Raise the Threshold for That Event
The larger the compensation, the less likely someone is to react immediately to a single shock event. They hold on longer. Part of that is the economics of a vesting schedule, but there's also a psychological pull: the more someone is paid, the more they're inclined to rationalize their own commitment to the organization.
Raising the threshold doesn't reduce how often these experiences occur — it just means people endure longer and then leave more abruptly. That's why someone holding a $1.1 million option package can seem to vanish out of nowhere. In reality, it isn't sudden at all. Events that had been accumulating for a long time simply crossed the threshold that compensation had been holding up.
For anyone running a small team, this mechanism carries a different lesson. The lower the pay, the heavier a single shock event lands, immediately. Having your contribution misattributed on a five-person team feels far more acute than the same experience on a team of fifty. The smaller the team, the more urgent it is to lower how often these experiences happen — before reaching for a bigger paycheck as the fix.
All Three Types of Shock Events Start With Conversations Never Had at Onboarding
Credit for contributions, decision-making authority, alignment on direction — all three failure modes trace back to promises that were never clearly made when someone first joined.
Misattributed contributions keep happening when there's no routine for recording who did what. If there's a shared, documented record of who made which decision and what came of it, attribution errors get caught and corrected quickly.
Erosion of decision-making authority usually happens because the scope of that authority was never clearly defined to begin with. Without an explicit promise up front — "this call is yours to make" — there's no basis for the person to judge, later, whether someone else stepping in counts as overreach or as help.
Misalignment on direction is the one that surfaces latest. Often, neither side can pinpoint when the gap started opening up between what the person wants to do at the organization and what the organization actually needs from them.
For all three, setting clear terms through a structured conversation within the first 90 days changes the intensity of whatever shock events follow. Corporate data analyzed by MIT Sloan researchers bears this out: teams that explicitly exchanged promises on contribution, authority, and direction during early onboarding show a different turnover rate a year later than teams that didn't.
Contribution Logs, Decision Rights, and Quarterly Check-Ins: How Small Teams Cut Down Shock Events
A contribution log can start far simpler than meeting minutes or a chat archive. Once a week, having each team member write two or three lines on the decisions they made and what came of them is enough. Simply having a format where contributions can be remembered and credited within the organization already lowers how often misattribution happens.
For decision-making authority, a simple list of "who has final say on this" is more useful than a job description. Hiring, pricing, client-facing calls, content direction — spell out the final decision-maker for each, and the odds drop that someone else stepping in later gets processed as a shock event.
Alignment on direction needs regular checking. Once a quarter, have a conversation to confirm whether, in the work they're currently doing, the person still finds a reason to keep working with this organization. If that conversation feels uncomfortable, the discomfort itself may be a sign that misalignment has already begun.
A $1.1 million option package is a scale no small team can match with pay. But cutting down shock events runs on a different track than compensation size altogether. If pay raises the threshold, lowering the frequency of shock events means reducing the number of incidents it takes to reach that threshold in the first place. The smaller the team, the faster promises about contribution, authority, and direction can be exchanged. What teams that keep turnover low actually do, when they can't compete on pay, is finish these three conversations early — in the first weeks on the job.
