"Historically grown." Do you know these two words? They usually come up when someone asks why something is so complicated. Then silence follows. They explain everything and change nothing.
In software development, this is called technical debt: decisions made in earlier projects, under different conditions, by people who have long since moved on. The code works, but nobody dares touch it, because nobody knows what else might break.
I know this from personal experience. Before I started advising companies, I wrote Access applications in Visual Basic for businesses myself. They worked, somehow, but nobody could explain them. Back then, I was part of the problem I describe today.
In companies, it is exactly the same. I call this «The Company Code».
🐛 The second code behind the org chart
Behind every org chart, process manual, and approval policy lies a second code. It consists of the actual dependencies and rules that are written nowhere, yet determine how decisions get made. Who knows which exception applies to which customer. Which approval is better obtained directly from the boss. Which table nobody is allowed to delete, even though nobody knows anymore what it is there for.
This code is not bad. It carries a lot of accumulated experience and makes a team adaptable. You see it most clearly when a situation arises that the manual does not cover, so someone decides pragmatically, and nothing gets written down. In software, this is called an uncaught exception: the programme keeps running, but something is accumulating in the background.
🐛 The number almost nobody knows
A useful measure for the unwritten code is organisational latency. Take a rule change from the last twelve months, for instance a new approval threshold or a changed discount. When was it decided? And when did everyone actually start doing it that way?
Almost nobody knows this gap for their own company. Yet it says more about an organisation's agility than any mission statement, namely how quickly it can actually change its own rules. How fast you adopt new technology is secondary by comparison.
🦋 Not everything needs writing down — just distinguishing
An objection comes up in almost every conversation: "That's exactly our strength. Our people think for themselves. If we write everything down, we lose that."
There is something to that. Some companies have documented their agility out of existence. The conclusion is still wrong, because you do not need to write everything down. It is enough to distinguish two things.
🦋 Operational knowledge belongs in the Company Code: who is responsible for which customer, which deadline applies, how a position is calculated. This is documentation that currently exists only in people's heads. It costs you time every day and makes you vulnerable when someone leaves.
🦋 Judgment stays with people: the sense that something is off in a project before the numbers show it. That cannot be written down, and it should not be.
Where the line between the two lies is a leadership decision. IT cannot take that away from leadership.
The new colleague doesn't ask
Why this gets more urgent with AI
Until now, the unwritten code was mainly a problem for new employees. They need months before they know how things "really" work. With AI agents, a new colleague joins who cannot learn the unwritten part at all, because what is written nowhere does not exist for them. More on this next week.
Wednesday, 8 am
Three steps to start
1️⃣ Have a process explained to you twice: once by the person responsible for it, once by someone who carries it out daily. Where the descriptions diverge, that is your unwritten code.
2️⃣ Measure a latency: one rule change, two dates, one gap.
3️⃣ Draw a line: what is operational knowledge, what is judgment?
Small steps beat no steps.
The unwritten code keeps running whether you read it or not. The only open question is who executes it next: a new person who asks, or an agent who does not.
👉 Where would you start in your company? We are happy to work through this together in a positioning session.
Part 2 of the series «The Company Code»: seven theses on how companies make AI effective without automating their chaos.
(AI modified)