“How many decision-makers does it take to make a call?” sounds like the setup for a cynical joke, or perhaps one of those brain-teaser interview questions from the 2010s where you’re supposed to multiply fifteen Fermi estimates and round the error terms generously. But I think it’s a fair question; turns out it’s not so hard to imagine possible worlds that require far fewer decision-makers than you’re used to. For example, a rock with a “don’t ever do anything” sign taped to it can have extremely high accuracy,Footnote 1Scott Alexander, Heuristics That Almost Always Work (opens in new tab) (archived Jul 2026). or, more relevant to this blog’s thematic bent, one could replace their central decision-makers with a bot that tweets “we leave implementation of $RECOMMENDATION to $SUBORDINATE_BODY” at random intervals.
Nondecision-making can feel abdicative, but much like the adage that not getting a response to your email is a response, there are a few good reasons to consider it here. One way to see it is along the lines of a blog post I wrote last year about the value of metaphorical inversions. In it, I used the inversion technique to reconceptualize governance as an “efficient, context-rich decision-scaling accelerator”. But here’s a new inversion: if that’s our definition, then why not choose the bot instead of the traditional governance body? It pre-empts all the usual anarchic decision-making failure modes: its behaviour is completely legible and therefore policy-compliant, it’s extremely lightweight, and the results are rational up to base rates. It’s the ultimate expression of the bias for action (opens in new tab), where the fastest way to deal with problems is to have pre-decided how to solve them.
You might protest that there’s no rational connection between the factual matrix surrounding a decision and the arbitrary phrases the Twitter account spews at 2am. But, as contrived and decoupled as its posts may feel, they actually reflect how many organizations do business, and for legitimate reasons. I recently wrote about how orgs sometimes engage in rituals — like the phenomenon of writing PowerPoint presentations that contain essentially no analysis — that appear irrational but are equally consistent with just having pre-cooked the strategy.It’s not unlike seeing culture as a set of first-order practices we tacitly agree to adopt for societal convenience. These organizations may be draping the problem in ceremony to confer legitimacy, in the sense that nobody will take your scribbled-on-a-Post-It strategy seriously.Footnote 2See e.g. Meyer and Rowan, Institutionalized Organizations: Formal Structure as Myth and Ceremony (opens in new tab) (archived Nov 2025), 1977. So the bot and the rock feel wrong because they’re not wearing any clothes, so to speak, but there’s nothing necessarily off about the results.
But even so, one problem with a well-tuned mental model for quick, consistent action is that your decisions might be indistinguishable from those of a rock or a bot. And this failure mode is invisible unless you’re willing to eat the cost of surfacing the counterfactual.Further evidence to support a judiciary slash strategy red team. Maybe you can’t run the analysis de novo every time, but you’ll always need a mechanism to signal when your core operating principles might be wrong. There’s definitely a cost here, but the punchline is that it doesn’t depend on headcount.
Two more types of nondecision
The classic nondecision framework is from Bachrach and Baratz’s Power and Poverty; it’s the one you feel whenever the meeting chair handles an issue by declining to put it on the agenda. But our analysis above suggests two more types of nondecision.
The first is motivated by Nils Brunsson’s work, which we looked at last time. His writing goes further than I did in my post though, noting in The Organization of Hypocrisy that many of the apparent hypocrisies inherent in org behaviour actually reflect a coherent strategy to manage complex relationships with disjunct audiences, each with their own irreconcilable demands. One way you’ll notice this, Brunsson contends, is that organizations that talk a lot about a given topic sometimes do very little.Footnote 3Nils Brunsson, The Organization of Hypocrisy, 1991, p. 28. Your strategy can’t make everyone happy, but if the way you talk about it does, then we’ve discovered another spin on interpassivity: the talk does the action so that we don’t have to. And on some level this just reflects our limited capacity to solve thorny problems; when you have competing priorities, something has to fall by the wayside. In the same way, your nondecision might be all talk.
But if the first nondecision mode is talk, then the second only sounds like talk; in reality it’s more of a way to manufacture the evidence required for later decision. Our indefatigable Twitter bot continues to delegate authorities, but while the text output might be memoryless, the system it interacts with isn’t. Eventually someone will notice if these powers are wielded inappropriately. And what then? The Twitter bot’s owning org will clarify their delegation. Critically this isn’t a new policy instrument, as that would pull the rug out from underneath the delegates and admit that the original declaration was inaccurate or defective. No, we’re clarifying the existing policy, which was always correct, so that our readers apply it correctly.
For example, imagine we needed to constrain our org’s branches in some deeply unpopular way. We could of course prohibit the action and resign ourselves to the ensuing reputational hit. But we could also allow the branches to handle the problem from the outset. They will almost surely deal with it in different ways, reflecting both individual differences in risk tolerance and the logical outcome of systemically vague power grants. And at an opportune moment we could simply release a statement that since the branches aren’t tackling the problem collectively, we’re obligated to clarify that the original delegation actually meant adopting the unpopular measure.
The outcomes are eventually equivalent, but along the way you’ll see high variability in action. Making sense of newfound powers probably requires a larger headcount, but the new employees’ work runs a higher risk of becoming overruled by the eventual clarification. So it ends up being a clever if less than forthright way of shaping organizational behaviour, but also an asymmetrically costly one.
The only way to win is to ignore what the bot said about delegation, and coordinate among branches from the start. And this is very close to another organizational contronym: the newfound power is less authority than obligation. And this means that if some projects really can get away with having one decision-maker, they’d better know when not to decide for themselves.