The Pyramid vs. The Chinese Room
The earliest grand structures people built, almost everywhere in the world, were pyramids.
Why? Was it . . . Aliens?
No, of course not. If you just pile bricks on top of each other, the natural shape they will form is a pyramid or mound. Organizational structures are the same way: As you pile new people into a company, it becomes a self-perpetuating, hierarchical pyramid. When a company is intentionally designed to not be a pile of people working together but instead a closed system of inputs and outputs, it can operate effectively with incredibly lean and modular operations.
How does one build the scaffolding for a corporate architecture like this? This can be done using what we call “Chinese Room” corporate architecture, a system we developed for managing our own network of companies (which at its height brought in around $70 million a year in top-line revenue with only a two person middle management layer).[[18]](#_ftn18)
The Chinese Room is a philosophical thought experiment in which an individual slips pieces of paper with Mandarin written on them into a room that subsequently produces thoughtful replies. In this room sits another man operating off of a set of instructions about how to reply to different things who does not speak or read Mandarin himself. The thought experiment asks whether or not we can say if the room “speaks Chinese.”
Chinese-Room-style corporate architecture involves learning to see a company as a black box with inputs and outputs, then isolating the functions required to transform the inputs into the outputs and dividing them into four categories:
- Can be performed by machine or script
- Must be performed by a human, but the human would not need to speak the language of the customer
- Must be performed by a human, but the human can be low-skilled
- Must be performed by a skilled human
If this mapping is done accurately, about 80% of a company or governance’s functions, especially middle management, should fall into categories one and two.
The trick is to see the organization’s human components as little different from its mechanical parts. As with our mechanical systems, humans process code we provide and require checks to ensure they process their code with fidelity. As with machines and programs, not all humans can run on the same code. These differences must be anticipated when writing code for an organization’s functions. Some people work better with written documents while others thrive with training videos combined with quizzes or mechanisms involving interpersonal and social interaction and reinforcement.
What allows this Chinese Room style of governance structure to be so much more effective than traditional governance structures is that it enables those operating the system to split test modules, capture institutional knowledge around what works, and replicate more optimal methods quickly. You may, for example, experiment with a few different scripts or training systems for a number of modules running the same task, determine which yields the best results, and proliferate that optimal “code” across all modules of that type.
Yes, having people operate according to their whims increases their efficiency slightly and can allow for breakout stars on your team—but we have never found or seen a way to successfully capture the unique talent of stars and transcribe said talent across a system. Instead, when filling roles for highly skilled people, we look for those whose natural whims, superpowers, and shortcomings fit perfectly for the given role. This often enables us to hire amazing people who are otherwise undervalued in the job market—perhaps because while they’re geniuses with the skills that matter for the role, they have health conditions or interpersonal social skills that remove them from contention for most mainstream jobs.
It is this modularized philosophy that allows organizations we have run to be lean, low cost, and efficient. It magnifies the impact A players can have while outsourcing roles that would have been held by B players. We were able to operate a team of hundreds of people living in five countries—without language overlap in many cases—to be managed from a single interface we built. We would love to see this methodology rolled out to larger governance structures but see it as unrealistic in the short term.
When a company using more traditional structures grows, it begins to suffer from bureaucratic bloat. The human pile of management hierarchy ends up increasing the cost of small tasks because their efficiency gets blunted by an ever-increasing number of people, departments, lawyers who demand oversight, managers who demand approval, HR managers who serve as middlemen, etc. A modularized system allows us to create a set of instructions that can be transcribed into as many modules as necessary, all running at optimal efficiency. This set of instructions can be constantly optimized (through split testing) and improvements can be rapidly deployed across the network of modules.