The corporate immune system does not reject new ideas. It digests them.

Freedom to think is not freedom to build

The edge is where unhinged creativity meets long-term goals. Every large organization says it wants one. Most offer the wrong freedom. You are free to think, to talk, to present. I have had all the freedom in the world to make the future sound exciting, and remarkably little freedom to build a piece of it. Beyond the slide deck, freedom thins out fast.

The classic argument for staying close to the core was muscle. Building took armies, and the armies came at a price: adapting to the immune system. That bargain is collapsing. With AI, a handful of people can build, test and run what recently required hundreds. Much of it no longer even needs a large cloud budget; more and more runs on local machines. The core's muscle is no longer worth the core's antibodies.

Permission is not freedom

This does not mean the edge should cut the cord. An edge without a connection to long-term goals, deep domain knowledge and real customers is not an edge. It is a hobby. The edge should stay connected while becoming steadily less dependent. Everything hinges on what the connection is made of.

Most connective tissue in large organizations transmits permission. Every new step needs a new yes, and the more visible you become, the more yeses you need. Success increases governance, so freedom shrinks as you grow. That is not a malfunction. That is the immune system working exactly as designed.

The connection worth building transmits freedom: access, trust and reach that expand as the edge learns.

My test is simple. Is your freedom growing or shrinking, month by month? If it is growing, you are on an edge. If it is shrinking, you are a department with a cooler name.

The value is not in the models

AI raises the stakes on all of this, because the value of AI does not sit mainly in the models. Models are becoming something everyone rents. The value sits in the layer above them: how the technology is orchestrated, put into context and tuned to the way people actually work. That layer cannot be built from a distance. It is built where the work happens.

One of my favorite examples is a Norwegian knitwear factory. The breakthrough was not a language model. It was machine vision watching the knitting machines for errors, explored by technologists standing next to the operators, with no translation layer of managers in between. Combined with new production technology, that direct line opened a path to bringing a fifth of offshored production back home. The innovation was not the algorithm. It was the organization: the shortest possible line between the person who knows the work and the person who can build. Every layer in between is immune tissue.

The immune system is learning to use AI too

Here is the uncomfortable part. Wire AI agents into the core, trained on the core's documents and rewarded by the core's incentives, and you do not get new knowledge. You get the old map amplified at machine speed: sycophantic confirmation, delivered instantly and in confident prose. The immune system gets an upgrade.

In a recent article in Magma I argue that leading AI agents is a task of epistemic design: deciding which roles agents play, where friction is built in, who revises whom, and how responsibility stays human. An agent can be designed to applaud your strategy or to attack its premise. Only one of those creates knowledge.

This is also why the edge ultimately protects itself with a loop, not a mandate. Build something small and put it into use fast. Let use improve the product, and let people feel the product improving because they use it. New knowledge emerges in the practice, not in the report. When that circle of interaction and data starts spinning, the edge draws its legitimacy from the outside in: from users and results, not from committees. The immune system can kill a business case. It cannot kill a product people love more every month.

A country has an immune system too

Zoom out, and the pattern holds. Norway is superbly organized, with layers, bodies and procedures for everything. That is our strength, and it is also our immune system. Fear of the future is rising here as everywhere else, and fear feeds the antibodies.

But we have placed outrageous bets before. Our heavy water sat at the center of the race for the atomic bomb, and in 1951 we were among the first countries in the world to run our own nuclear reactor. Two decades later we drilled through the seabed and built a fortune. I think of AI as both bets in one: the bomb and the oil, power and wealth in the same technology.

Bets like that are not placed by procedure. They are placed in environments where thinkers and doers share a room, where young energy is let in, and where the connection to our large institutions carries freedom, not just permission.

An idea the immune system cannot digest has a name. We call it the future.