Intelligence is scaling fast. Judgment isn't.

We are used to treating intelligence as a highly valuable resource in any organization. We hire for it, pay a premium for it, and build entire strategies around access to it. That assumption is now breaking. For the first time, intelligence scales with something other than people. It scales with compute, and compute is being built out at a pace McKinsey expects to reach seven trillion dollars by 2030.

Traditional media is having a field day with the anxieties this creates. AI will gut white-collar work, hollow out human relationships, perhaps even end the world itself. Some of it may even turn out to be true. Nobody knows. But before deciding what we are up against, we should decompose what we actually mean by intelligence.

What AI is providing in its current form is a specific kind of intelligence: raw processing power, pattern recognition, information synthesis. What economists have called the collapsing cost of prediction. When we conflate that narrow capability with intelligence in the full human sense, we end up concluding that machines can replace humans wholesale. The more useful question is not how close machines are to us. It is where value moves when one layer of intelligence becomes abundant. To answer that, it helps to lay intelligence out as a hierarchy. Five levels, with machines playing a different role at each. Stacking intelligence this way is not a new idea. T.S. Eliot was asking "Where is the wisdom we have lost in knowledge?" in 1934, and management thinkers have been drawing data-to-wisdom hierarchies since the 1980s. What is new is that machines now occupy a growing piece of the lower levels.

The hierarchy: from raw input to wisdom

Level 1: Knowledge

At the base, intelligence is knowledge: the accumulation of facts, information, and skills through experience and education. It answers the question "What do we know?"

Every major technology shift has profoundly changed the role of raw human knowledge. Consider the printing press, the calculator, and even the internet. Each reduced the value of remembering things and freed humans to approach problems in new ways. In the AI era, knowledge can (somewhat reductively) be defined as training data. Access to it used to be a massive competitive advantage. The internet and LLMs have largely leveled it.

Level 2: Intelligence, narrowly defined

One level up in this hierarchy sits intelligence in the narrow sense: the ability to process information, recognize complex patterns, solve problems, optimize under constraints, and learn. It answers the question "How do I solve this problem?", drawing on the knowledge below it.

This is the definition at work in the term artificial intelligence. AI can write code, summarize long documents, and find anomalies in medical scans in seconds. But this kind of intelligence is instrumental: it figures out how to achieve a goal efficiently. Crucially, it does not choose the goal.

Level 3: Understanding

If processing were enough, every problem could be solved by a big enough calculator. Anyone who has actually run a company knows it does not. Understanding is the capacity to grasp underlying meaning, causal relationships, and context. Its the bridge between raw intelligence and judgment. It answers the question "Why do these facts matter in this specific situation?"

An LLM currently simulates understanding remarkably well by predicting what plausibly comes next. But most of the context that determines whether an answer is right for you exists nowhere in writing: the dynamics in the room, what happened the last time you tried this, or the customer data point there is no CRM field for. A model cannot weigh context it has never seen.

Level 4: Judgment

Once we understand, we can prioritize. Judgment is the ability to weigh competing priorities, evaluate trade-offs, apply values, and make a definitive choice where there is no mathematically correct answer. It answers the question "Given the options, what should we do?", and it lets us act decisively even when our knowledge, intelligence, and understanding are incomplete. They almost always are.

Understanding can give you three optimized routes to a destination; judgment decides which one to take, based on your risk tolerance or your ethical boundaries. And judgment has a property the lower levels lack: it is not just a cognitive act but a social one. A decision needs an owner. AI can generate a decision, but it cannot own the consequences. It has no skin in the game, and your board, your bank, and your employees will not accept "the model decided" as an answer. That accountability is precisely why judgment commands a premium.

Level 5: Wisdom

Most of us have met people who seem to make the right call even on incomplete information. But there is a level beyond good judgment, and we know it when we see it. Wisdom is the synthesis of all the levels below, plus long experience: a deep perception of human nature and systemic consequences. It is knowing the difference between what you can do and what you should do, over a long time horizon. It answers the question "What is this choice ultimately worth, for the people it touches?"

Whatever the AI models learn to imitate, wisdom is earned, not computed. It requires a lifetime of lived stakes, mistakes, and emotional maturity.

When intelligence becomes free, the bottleneck moves

If anyone can generate highly intelligent solutions instantly, the scarce and valuable levels become the ones above.

We have been through this migration before. The internet moved the bottleneck from knowledge to processing: from knowing things to doing something with what everyone could suddenly know. AI is now moving it from processing to understanding, judgment, and wisdom. We are not being replaced; we are being pushed another step or two up a ladder we have been climbing for a long time.

What this means for a Nordic SME

Strip away the Silicon Valley abstraction and look at an ordinary decision. Say you are considering expansion into a new country. AI will hand you the market analysis, the competitor landscape, and three entry strategies fairly instantly. It will provide you with knowledge and intelligence, and do so effectively for free. This is in itself revolutionary. That said, there are many things it cannot provide you. For instance the emotional state of your team, whether you can stomach eighteen months of losses to get there (judgment), or whether the company you would have to become is one you actually want to lead (wisdom).

The strategy deck or the code was never actually the hard part. It is just the part that used to be expensive and require a special kind of access to particular knowledge. That access is now ubiquitous.

So no, AI is not replacing human input. It is repricing it. Knowledge and processing is quickly becoming cheaper; understanding, judgment, and wisdom are heading the other way. The leaders who come out of this decade ahead will not be the ones with the smartest tools. Everyone will have those. They will be the ones with the judgment to decide what the tools are for.

P.S. This is a deliberately simplified framework, and it leaves things out. Two omissions worth naming: intuition — the brain's ability to decide from thousands of past experiences without calculating the steps, tied to emotion and gut feeling in a way statistical pattern-matching is not; and values — the belief system that defines what is "good" and "just" in the first place. Without values, intelligence has no direction to optimize toward. Both deserve essays of their own.