Residual Judgment Right: Why Organizations Ultimately Need Someone to Call the Shots

In this session, we delve into the “Companies and Leadership” module. Before we discuss specifically how to lead, let’s first consider a more fundamental question: whether it’s a company or an organization, why must there be a leader?

Many people believe this isn’t a problem at all. As the saying goes, “a household, no matter how large, needs a single head,” so a company should listen to its boss. An autocratic CEO, they believe, should be able to make strategic decisions at the top, guide designers on color schemes in the middle, and supervise cafeteria cooking at the bottom.

However, some people don’t see it that way. In this day and age, who is significantly smarter than anyone else? Shouldn’t resources be allocated by the market, answers provided by data, and solutions analyzed by AI? When indecisive, why not hold a democratic discussion or have everyone vote, instead of having to obey one person? Is this scientific?

We will derive from first principles why an organization ultimately must have one person—a specific individual—in charge.

The result of our deduction is called the “residual judgment right.”

Residual Judgment Right: The Cornerstone of Economic Theory #

What is obvious to the common person and to a boss is not so obvious to scholars. Let’s first see how scholars convinced themselves.

Before you can argue that a company needs a boss, you first need to prove why companies are needed in the world. Since the market is so efficient, why not let everyone do whatever they want, and achieve division of labor and cooperation through market transactions? Why must companies be formed, with everyone going to a fixed workplace every day, confining themselves?

In 1937, British economist Ronald H. Coase found the answer: “transaction costs” [1]. To participate in the market, you have to search for prices, negotiate, sign contracts, monitor performance… all of which incur costs, both monetary and time-based. You must free yourself from these administrative tasks to focus on actual work.

And as long as a company is established, people can easily divide labor and cooperate within the company: if you want to use certain resources or people, you just use them; there’s no need to renegotiate prices every time.

But then you might say: I understand establishing a company, but I don’t understand why I must obey the boss’s command. Before joining a company, I could perfectly well sign a contract specifying what I’m supposed to do, right? Why should I have to obey his commands like a beast of burden if the boss suddenly has a whim?

In 1951, American economist and management theorist Herbert A. Simon explained why authority exists within a company [2]. When a company hires an engineer today, it’s impossible to specify in the contract every single line of code he will change over the next three years. Therefore, employment contracts differ from ordinary sales contracts: what the boss buys is not a fixed task list, but the right to direct within defined boundaries.

Simply put, Simon proposed that the boss must be able to direct employees at will within a certain scope. When employees sign an employment contract, they are essentially agreeing in advance to let the employer decide what they specifically do within an “area of acceptance.”

By the 1980s, American economist Sanford J. Grossman and British economist Oliver D. Hart further developed the theory of “incomplete contracts” [3]: the future cannot be exhaustively written into a contract, therefore, the truly important part of asset ownership is precisely the “residual control rights” over matters not specified in the contract.

In other words, what does it mean for a boss to own a company? Ownership means: for critical assets whose use is not stipulated in the contract, I have ultimate control, and this asset control in turn determines my organizational decision-making power—you must listen to me on matters not explicitly written down. Otherwise, if everyone just follows the contract, where does my “power” manifest itself? A boss is not merely someone who enjoys dividends; a boss is someone who calls the shots.

In this way, discretionary power converges from employees to the boss. But even if the boss has the right to decide, what exactly should he decide?

This is where American economist Frank H. Knight, whom we have mentioned many times, comes in. Knight said that the boss’s decisions are about how to deal with uncertainty.

We previously discussed the distinction between “uncertainty” and “risk”: risk can be quantified with probabilities, and you can always use some method to hedge against it (e.g., buying insurance), thus dealing with it rationally; uncertainty, however, doesn’t even have a probability distribution. You don’t know what will happen in the future; it’s likely something no one anticipated today. Therefore, this matter cannot be treated rationally; it requires a determined gamble.

This type of product has never been seen on the market, and no one knows if consumers will buy it. Should we develop it? How much are we willing to invest upfront? What should our production scale be? No matter how much rational component these decisions have, there will ultimately be an irrational, willful element—what we’ve previously referred to as “inductive bias” and “subjective prior”—this is the uncertainty the boss must bear. Knight said that the boss’s profit is compensation for bearing this uncertainty [4].

Coase discovered transaction costs → Simon discovered the area of acceptance → Grossman and Hart discovered residual control rights → Knight discovered judgment under uncertainty.

Thus, we’ve deduced from “companies must exist” to “someone must be in charge of a company.” But why must this decision-maker be a single person, and not a committee or even an AI?

Why Must One Person Ultimately Make the Final Decision? #

Firstly, a boss cannot merely say, “I’ll just act according to laws and regulations, and make no subjective decisions.”

Laws and rules can only define the scope of decisions—for example, you cannot commit fraud, your budget cannot exceed a certain amount, and what safety standards are—but they cannot answer whether we should actually make this product, or whether we serve high-end or mass-market customers. If we encounter short-term losses, should we persevere or retreat? This is like how the laws of physics can rule out houses that cannot be built, but cannot automatically generate a house worth building from all possible architectures.

Furthermore, nor can a boss say, “I’ll listen to the committee; we’ll make democratic decisions on issues.”

Voting allows everyone to express preferences and can form a majority decision for each option, but piecing together what everyone likes best cannot be called a strategy. For example, you could organize a vote on product positioning, then another vote on the technical roadmap, then vote to decide pricing, then vote on brand image… The final result you get will be a “frankenstein”:

The positioning is high-end, the technology is simple, the pricing is affordable, and the brand is luxurious.

These four requirements are fundamentally incompatible, aren’t they?! There’s a mathematical principle at play here. American economist Kenneth J. Arrow proved [5] that as long as there are more than two options, there is no universal voting method that can guarantee combining everyone’s preferences into a consistently rational and coherent collective choice. This is known as “Arrow’s impossibility theorem.”

In other words, voting cannot guarantee you a strategy. Strategy requires you to redefine problems, create new options, and must ensure that product, technology, brand, pricing, and organizational structure form a coherent set. Strategy isn’t assembled from individual preferences.

Then you might ask if we can rely on “self-organization of complex systems,” allowing various departments of the company to freely explore directions, combine freely, and eventually have the best decision emerge? No, that won’t work either. Self-organization is only suitable for tasks with high modularity, low trial-and-error costs, and where errors can be locally isolated [6].

Everyone can write software modules independently, but someone ultimately needs to decide which code can be merged into the main version. Grasslands can grow naturally, but the two ends of a bridge cannot spontaneously meet in the middle of a river through “emergence.” Someone always has to issue a definitive command.

What about using AI for decision-making? That won’t work either. It’s important to remember that AI’s output depends on the goals, constraints, and evaluation criteria you set for it: it can tell you the potential costs of various options, but it cannot decide for you what is worth paying for.

Ultimately, no matter what procedure is used or how much discussion takes place, a definitive decision must ultimately be made. A jury can argue furiously behind closed doors, but when they emerge, they can only deliver one verdict: guilty or not guilty. It cannot present a report to the judge saying, “Sixty percent of people think the defendant is guilty.” An organization cannot ultimately deliver “what everyone thinks”; it can only deliver “what to do now.”

A company must, when evidence is insufficient and outcomes are yet unknown, choose one from several incompatible futures and stake its capital, time, talent, and reputation on it.

This indivisible critical gamble can only have one ultimate master. Of course, it doesn’t necessarily have to be a flesh-and-blood person; it could also be a “single-valued final mechanism”… but it must have a clear value function and accountability, it cannot be one way today and another tomorrow, and it cannot merely say “I’m just helping with ideas”—which is why it’s usually a single person.

Leadership is the system that compresses “many possibilities” into “one shared commitment.”

An organization must have one person who, after incomplete contracts and facing an uncertain future, sets a subjective strategy, ensures its implementation, and is willing to bear responsibility for it.

This is the residual judgment right.

The Boss’s Scope of Responsibilities: What to Manage and What Not to Manage? #

Then you might say, after all this lengthy deduction, we’ve simply proven that the boss should call the shots. Isn’t this common sense for ordinary people? What’s the point of this? The significance is immense.

The key word in “residual judgment right” is “residual.” A boss cannot manage everything.

The standard for who should manage an issue depends on its “coupling” and “reversibility”: coupling refers to whether a decision will involve many parts of the organization, requiring them to coordinate with each other; reversibility refers to whether a decision, if wrong, can be withdrawn at a relatively low cost.

Using these two dimensions, we can divide an organization’s affairs, big and small, into four categories—

Matters with low coupling and high reversibility should be delegated to the people on the ground. Whether customer service waives a shipping fee, what color a page button should be, how a small feature is implemented—these decisions are best made quickly, and even if they go wrong, there’s not much loss. Avoid any approval processes for them.

Matters with low coupling and low reversibility should be entrusted to experts and clearly designated responsible persons. Accounting treatments, security protocols, encryption schemes—these cannot be handled however anyone pleases; they must be reliable, and someone must be accountable. However, the boss cannot make them more reliable; the advice of high-ranking professionals should be followed.

Matters with high coupling and high reversibility will affect the entire company but not cause any major harm. These can be delegated to department heads for independent exploration, but the boss must oversee the overall picture. For example, you can set up several teams to test multiple product prototypes simultaneously, but you must ensure that the infrastructure used by each team is compatible.

As for matters with high coupling and low reversibility, because they have a wide impact and can lead to significant losses if problems arise, they must be personally decided by the boss. Company positioning, core architecture, major mergers and acquisitions, factory site selection, key personnel decisions, and whether to self-disrupt all fall into this category.

When discussing options earlier, we mentioned Bezos’s “two-way door” and “one-way door” concepts: reversible decisions are two-way doors, while significant and hard-to-reverse decisions are one-way doors. It’s clear, then, that the boss’s most crucial task is to make those one-way door decisions. Bezos’s warning, however, is that the most common ailment of large companies is using heavy, one-way door processes to handle a large number of two-way door decisions, which makes organizations increasingly slow and reduces experimentation.

Some bosses like to directly intervene in frontline work, which management calls “micromanagement” (or “micro-operating”), and that’s not a good term. When you reclaim judgments that should be made on-site based on local knowledge, it may seem responsible, but in reality, you are weakening the organization’s judgment capability.

A boss should absolutely not manage more than necessary. Anything that can be codified into a process, delegate to processes; anything that can be determined by price, delegate to the market; anything that can be determined by professional knowledge, delegate to experts; anything that allows for quick trial-and-error with negligible consequences, delegate to the frontline… Delegate everything that can be delegated. The small remainder—the part that cannot be further proceduralized, marketized, specialized, or localized—those highly coupled, difficult-to-reverse, insufficiently evidenced, and time-sensitive problems, those are what the boss should truly manage.

Yet, the domineering CEO scenarios most often fantasized about by the public are precisely where power is used on trivial matters.

Five Core Judgments of a Leader #

Starting from the concept of residual judgment right, we can deduce that there are five things a leader should genuinely decide themselves—

First is problem definition, that is, what problem the organization is truly solving.

Is a sales decline due to underperforming sales staff, or has the product lost its value? Is a project delay due to poor execution, or are the objectives themselves conflicting? Is it that employees lack a “wolf spirit” (aggressiveness), or do they simply not know what winning means?

You must find the correct narrative.

Second is setting the measure, that is, defining the objective function.

Profit, growth, quality, speed, resilience, customer benefit, and employee development—these are all good things, but they often conflict. You must decide their priorities, what are the hard constraints that cannot be broken, what can be temporarily sacrificed, how to define winning, and over what time scale to evaluate results.

Third is making the bet, that is, strategic trade-offs.

As we’ve discussed, strategy is not just about what you want, but more importantly, what you are willing to give up for it. Can you ensure consistency in strategy when facing future difficulties?

Fourth is delegation, that is, deciding what matters are no longer decided by yourself.

Local knowledge resides at the frontline, but global externalities are at the top. You need to decide who holds proposal rights, approval rights, execution rights, and veto rights, what situations require escalation, and how bad news can bypass layers of embellishment to reach the center of power.

Fifth is re-evaluation, that is, deciding what evidence necessitates admitting a mistake.

You must be both resilient and able to know when to turn back after hitting a dead end. So how do you know if you’ve hit that wall? You must think beforehand which signals prove your strategy effective and which signals prove it ineffective. True steadfastness means abandoning old answers for the sake of the goal; false steadfastness means sacrificing the goal to protect old answers.

A Leader’s Daily Workflow and Future Outlook #

Seen this way, a leader’s daily workflow is simply—

Listen to information and suggestions from all parties →

Choose a future, define the strategy →

Narrate, helping people understand why this future is chosen →

Design an incentive-compatible system, turning the company into a machine to achieve your chosen future →

Listen to feedback, ensuring your team has psychological safety →

Correct errors.

I feel all of this is very much like harnessing a group of AI agents for a project. Even if you’re not a leader today, you still need leadership; at the very least, you need to lead AI.

Even if you are being led, you should still know how an organization can operate efficiently. We’ll delve into this more later.

【As the Gatha Attests】

Plans may gather collective wisdom, but decisions must rest on one’s shoulders; Rules delineate boundaries, but a path must be chosen from diverging roads. A hundred strategies can be simulated by machine, but success or failure rests on human responsibility; It’s not to boast of one’s own wisdom, but because accountability must have a clear locus.

Notes

[1] Coase, Ronald H. “The Nature of the Firm.” Economica 4, no. 16 (1937): 386–405. Coase originally referred to this as “marketing costs”; “transaction costs” is the name later given to this idea.

[2] Simon, Herbert A. “A Formal Theory of the Employment Relationship.” Econometrica 19, no. 3 (1951): 293–305.

[3] Grossman, Sanford J., and Oliver D. Hart. “The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration.” Journal of Political Economy 94, no. 4 (1986): 691–719.

[4] Knight, Frank H. Risk, Uncertainty and Profit. Boston: Houghton Mifflin, 1921.

[5] Arrow, Kenneth J. Social Choice and Individual Values. 3rd ed. New Haven, CT: Yale University Press, 2012. Originally published 1951.

[6] Benkler, Yochai. “Coase’s Penguin, or, Linux and the Nature of the Firm.” Yale Law Journal 112, no. 3 (2002): 369–446.