Lean, Six Sigma, Agile, and now: managing intelligence

Every era of business has had a discipline that defined how good organizations operated. Not a tool and not a vendor, but a way of thinking that separated the companies in control of themselves from the ones that were not.

Lean brought the management of waste. Six Sigma brought the management of variation. Agile brought the management of change. A fourth is now forming, and its subject is intelligence: how an organization turns what it knows into decisions it can act on. Call the era AI Native. The discipline underneath it is the management of intelligence.

Four successive eras, each with its own defining discipline.

It is tempting to read these four as one tradition that evolved in a straight line. They did not. Lean and Six Sigma grew out of manufacturing. Agile came out of software. The management of intelligence grows out of data. They are not one bloodline. They are the defining discipline of four successive eras, which is the stronger claim, because it means the pattern holds across completely different terrain.

And there is a pattern. It is the part almost everyone misses, because it sits underneath the part everyone remembers.

The famous half and the hard half

Each of these disciplines is remembered for how it manages an output. Waste, variation, change, intelligence. But none of them ever worked on the output directly. Each one first imposed rigor on the layer beneath it, and that unglamorous substrate work was the actual discipline. The famous half got the name. The hard half did the work.

Every famous output rests on a substrate that made the discipline work.

Lean is remembered for eliminating waste. But you cannot eliminate waste you cannot see, and most waste is invisible until the process is made visible. So the real work of Lean was never the elimination. It was standardized work and value-stream mapping: turning a vague, variable, tribal process into a documented standard you could actually look at. Once the process was visible and agreed, the waste stood out on its own. The substrate was a disciplined, visible process. Remove that, and Lean is just a poster on the wall.

Six Sigma is remembered for reducing variation to almost nothing. But before you can reduce variation, you have to be able to measure it, and trust the measurement. This is why the M in DMAIC comes before the I, and why measurement systems analysis exists at all: a Six Sigma project spends real effort proving the gauge is reliable before it touches the process, because reducing variation you cannot trust the reading of is worse than doing nothing. The substrate was trustworthy measurement. Remove that, and every improvement is built on a number that might be noise.

Agile is remembered for managing change through short iterations. But you cannot steer by change you cannot see quickly and honestly. So the real discipline was the feedback loop: working software at the end of every sprint as the single, current, undeniable signal of where things actually stand. Documentation could lie, plans could drift, but running software at the end of a sprint could not. The substrate was an honest, current signal of state. Remove that, and Agile becomes a set of meetings.

In all three, the same shape. The output that made the discipline famous rested on a substrate that made the discipline work. And in all three, the organizations that skipped the substrate got the vocabulary without the result.

The substrate of intelligence is meaning

The management of intelligence is no different, and its substrate has a name too. It is meaning.

For an AI to turn what an organization knows into a decision it can act on, the knowledge underneath it has to be disciplined first. That means definitions that hold across the whole organization, so revenue is one thing and not four. It means data you can trust, with the reliability of each source made explicit rather than assumed. It means ownership that is clear, so every number has someone accountable for it. And it means governed context an AI can reason over without inventing what it does not know.

Each of these maps cleanly onto a substrate discipline that came before it. A shared definition is this era's standardized work: the agreed standard that makes the rest legible. A data contract is this era's measurement systems analysis: the explicit guarantee that the reading can be trusted before anything is built on it. Clear ownership is this era's process owner: the named accountability without which a standard quietly decays. The management of intelligence is not borrowing the prestige of these disciplines. It is doing the same thing they did, on the substrate that matters now.

What skipping the substrate looks like

The failure is easy to recognize once you know the shape, because it is happening in most organizations that reached for AI first.

A pilot is built. It demonstrates beautifully on a curated slice of data, and then it meets production data and falls apart, because the curated slice hid every disagreement in the underlying meaning. Two departments ask the same question and get two answers, because each defines the key term differently and neither knows the other does. The model, asked to reason over meaning that was never agreed, fills the gap the only way it can, by inventing a definition that sounds plausible. And at the end sits a number that demonstrates well and that no one will sign, because no one can trace where it came from or defend it in front of a board.

A substrate failure, not a model failure: unagreed meaning produces two answers to one question.

None of this is a model failure. The model is doing exactly what it was asked. It is a substrate failure, and it is the direct, predictable result of managing the output before disciplining the layer beneath it. Every era had its version of this. This is ours.

Managing intelligence as a practice

Stated as a sequence, the discipline is almost boring, which is how you know it is real. Get the processes straight. Agree the definitions. Make the data trustworthy, with its reliability explicit. Only then apply intelligence on top.

The order is the lesson: discipline the substrate before you manage the output.

A quality engineer will recognize that order immediately, because it is the same order every operational discipline has insisted on. Define before you measure. Measure before you improve. Discipline the substrate before you manage the output. The sequence is not a matter of taste or a nice-to-have that a fast team can skip under deadline. It is the entire lesson, compressed: the output cannot be managed until the substrate has been disciplined, and every attempt to reverse the order produces the failure above.

This is the discipline Datahub is built to practice. The management of meaning, so that intelligence can be managed on top of it.

The objection worth answering

A careful reader will push back here. Is this not just data governance with a new name?

It is not, and the distinction matters. Governance disciplines the substrate. It is necessary, and it is exactly the unglamorous work described above. But a perfectly governed data warehouse does not, on its own, turn what an organization knows into decisions it can act on. Governance is the substrate discipline. The management of intelligence is what you do with the output once that substrate is trustworthy. Governance is necessary and not sufficient, in precisely the way trustworthy measurement is necessary and not sufficient for Six Sigma. Confusing the two is how organizations end up with immaculate catalogs that no one uses to decide anything.

The other reasonable objection is that AI Native names a state, not a practice, in the way cloud native does. That is fair, and it is why the era and the discipline are worth separating. AI Native is the era. The management of intelligence is the discipline practised within it, and it is entirely concrete: definitions, contracts, ownership, and reasoning built on top of them.

The point

The management of intelligence is not a break from Lean, Six Sigma and Agile. It is the next entry in the same logic, applied to the substrate that matters now. Same shape, new terrain. The unglamorous work, done first, so the output can be managed at all.

Every era learned the lesson the hard way, by watching the organizations that skipped the substrate get the vocabulary and none of the result. This era will learn it too. The only question is whether an organization learns it before or after its first AI project produces a number no one will sign.

The substrate has a name. Discipline it first.

Meaning.

Further reading: the mental model under this discipline is worked out in The company brain: where your organization's knowledge lives.