The Human Code Was Never About AI

When I started calling my work The Human Code, people assumed it was a response to AI.

They were wrong.

The Human Code was never about AI. It was about what AI keeps reminding us we already had and forgot to name.

When I took my first HR role over two decades ago, the smartest people in every organization I worked in were the ones holding the parts of the system nobody had written down. The manager who read a room before the first slide came up. The executive assistant who knew which meetings were really about something else. The senior HR partner who could tell you which department was about to lose two people without looking at an engagement survey.

None of them were doing magic. They were running a kind of internal operating system that had never been mapped. Judgment. Pattern recognition. Institutional memory. Discretion. The quiet work of knowing what cannot be said yet, and knowing when to say it.

Organizations used to call this experience. A few still do. Most have quietly stopped.

What AI is actually revealing

I have spent the last three years watching senior leaders at every level wrestle with the same question. What do we do when AI can write the policy. What do we do when it can run the screening. What do we do when it can draft the performance review, pick the candidate, recommend the reorganization.

The question they keep asking is what AI will replace.

The question I keep asking is what AI is making visible.

Because the thing AI is making visible is this. Most organizations have been running on uncodified human code for their entire lives. The performance review process looks like a system on paper. In practice, it runs on a manager who knows how to pull a difficult conversation out of a five-word Slack message. The leadership pipeline looks like a framework. In practice, it runs on three people who have been mentoring under the radar for eight years.

When those three people leave, the pipeline does not stop working because the framework broke. It stops working because the code was in them. And nobody ever wrote it down.

What the code actually is

The Human Code has five components. I have spent two decades watching each of them go unrecognized, unpaid, undocumented, and unreplaceable.

Judgment. The non-protocol decision. The moment where the rule says one thing, and the person looking at the situation knows that following the rule here will cost more than it saves. Every organization runs on these calls. None of them write the calls down. The call looks like experience. It is actually a compressed pattern recognition engine that someone built slowly, in public, usually without anyone acknowledging the cost of the building.

Relationship. The thread of trust between people who have never been introduced on an org chart. The manager who will tell you the truth because you hired her predecessor. The finance partner who will flag the budget risk before it hits the dashboard because you remembered her son's name. Organizations move at the speed of these threads. They fracture at the speed of ignoring them.

Memory. The institutional why. The reason the policy exists. The reason the reporting line is structured this way. The reason that conversation with that director was ended in that particular way in 2019. AI can query your policy library. It cannot answer why the policy was written. The answer is in someone's head. Usually someone senior. Usually someone close to leaving.

Discretion. The weight of what you carry silently. The announcement that is still three weeks out. The name on the layoff list that you cannot yet say. The peer whose performance is an open problem but who still has four months of runway. Discretion is not optics. It is a structural load-bearing element of any organization that moves people for a living.

Integration. Knowing how the pieces fit without the map. What a handoff between two departments actually costs in hours and attention. Which VP will sign off on a proposal only if you walk the language backwards from his deputy first. Which Slack channel sets the real tone for a week. Integration is the part of the work that has never once appeared on a job description, and the entire organization relies on it.

If this essay named something you have been sitting with, subscribe. Beyond the Title sends one essay a week on the architecture of work.

AI is the mirror

I have started telling the senior leaders I work with that AI is not the replacement event. AI is the mirror event.

AI is holding up, clearly for the first time, the shape of what an organization has always been running on but has never been asked to name. And the shape is uncomfortable to look at. Because the shape says that most of the function that organizations call infrastructure was in fact held in a small number of people, quietly, for a long time, without a plan for succession or documentation or acknowledgment.

That is the news. It is not that AI will take these things. It is that AI cannot take them, and the fact that AI cannot take them is finally making it obvious how much organizations needed them in the first place.

There is a version of the next five years where senior leaders say that the quiet parts of the work were always the loud parts. That what got called soft was in fact load-bearing. That the manager who knows how to hold an uncertain room was the entire reason the operating model functioned through two restructures and a pandemic. And that if you do not know how to value this work, you are going to spend the next cycle paying to rebuild what you never knew you had.

There is also a version where organizations pretend the mirror is not there. Where they automate the documentable parts, call it a transformation, and lose the undocumented parts one senior departure at a time. That version ends badly. It has already ended badly for a few companies I can name privately but will not here.

What the code is actually for

Here is the part I do not often say aloud.

The Human Code is the reason the work matters.

If an organization could run without judgment, without relationship, without memory, without discretion, without integration, it would not actually be an organization. It would be a workflow. The people inside the organization would sense the difference, even if they could not name it. The trust that holds large enterprises together, the thing that lets a new hire believe that the company will handle a difficult situation with care, does not come from the framework. It comes from the five things that AI cannot hold.

The fact that those five things are heavy to carry is not an accident. It is the signal that the work is being taken seriously.

I have stopped trying to replace them with systems. I have started trying to document them where I can, protect the people who carry them, and build deliberate architecture around the parts that cannot be written down.

That is the work of the next decade.

What I have learned

Twenty years in, I can say this with confidence.

The Human Code was always the point. AI did not create the question. AI just made ignoring the question impossible.

The organizations that will do well in the next decade are the ones that look at what has been holding them together, name it, document it where they can, and build deliberate infrastructure around the parts that cannot be written down.

The organizations that will not do well are the ones that look at AI and ask only what they can remove. They will remove the wrong things. They already are.

I wrote the book because I wanted to name the code before another generation of senior leaders retired without anyone sitting with them long enough to transfer it.

What does your organization run on that nobody has ever been asked to name?

Roy Tran Beyond the Title