Signals

For four centuries the way to build a flying machine was obvious: copy the bird. The results were not encouraging. What that failure teaches about designing work between people and AI.

Organizations spent decades eliminating variation. Generative AI produces it by design. Knowing which conditions you are operating in decides whether control helps or hurts.

Every book written about AI is out of date by the time it is printed. The static manuscript was the right container when the constraint was distribution. That scarcity has inverted.

Every enterprise runs one system that tells it who its people are. That system was built on a philosophy older than anyone running it. HCMs will not be reinvented. They will split.

If the machine and I now write the same way, whose writing is it? An experiment in asking Claude to confess to its own style, and then take the confession back.

The most-quoted number in enterprise AI comes from a single yes-or-no question. Almost nobody who quotes it has read the appendix. It is a rumor with a decimal point.

Work was always doing two things at once. It produced output, and it developed capability. Organizations optimized for the first and assumed the second. That assumption no longer holds.

The mirror image of the obsolete list. When AI makes judgment the scarce resource, the practices that govern judgment become the most important work in the building.

For a century, being a late adopter was the safe bet. The one condition that made it safe — that competitive advantage plateaus and diffuses — is gone.

The first question of the AI era is not which jobs a machine can take. It is what the work should become. AI does not do human work faster — it does different work to reach the same outcome.

Organizations treated the human layer around AI as friction. Commonwealth Bank and Ford show what that layer costs when it is removed and has to be bought back.

AI does not reform human resources. It inverts it — takes the function’s purpose and turns it into its opposite, while leaving it in the same seat. The name on the door is the last thing to change.

Most HR methods transform. A few are elevated. These ten simply end — because they solve a problem that no longer exists. They don’t need to be redesigned. They need to be retired.

Management theories don’t fail because they were wrong. They fail because the world they were right for changes. AI is changing it again — as profoundly as World War II did.

Bureaucracy gets a bad name, but it solved a real problem. AI changes the underlying economics — the new constraint is human agency, and bureaucracy was never designed to maximize it.

Two organizations can begin with the same access to intelligence. Within a year, one is measurably ahead — not because the technology changed, but because of what happened every time someone had to exercise judgment.

The industrial era struck a deal with workers: give us your craft, and we'll break it into tasks anyone can do. AI reverses that bargain — and most organizations are still optimizing the old deal.

There is a pattern emerging in organizations serious about AI: people are spending more time checking outputs, validating recommendations, applying judgment. That interpretation gets the design exactly backwards.

Most organizations measure return on AI the same way they measure every other investment: cost reduction, headcount eliminated, tasks automated. This is scarcity logic applied to an amplification technology.

In 1960, McGregor changed how the world thinks about management. Then he tried to apply Theory Y inside an architecture that was still fundamentally Theory X. The structure won. It always does.

Not every company can transform in place. That’s the honest answer nobody puts in the consulting brochure. Most approaches stall at exactly the same point.

The most natural thing in the world is to look at what your people do and ask, “Can AI do that faster?” It’s also a trap — when you automate existing tasks, you’re assuming those tasks should exist.

When organizations deploy AI to replicate human tasks — faster, cheaper, at scale — they are building robotaxis. Systems that perform brilliantly within their training and freeze the moment conditions change.