What is Industry 4.0?
Industry 4.0 is the idea that an operation can sense, decide and adjust itself, rather than being run entirely by people reading instruments and turning things.
The useful way to understand it is through the three revolutions before it. Each arrived when a new foundation made a new kind of operation possible: power, then standardisation, then computation, and now context.
The first: machines replace muscle
From around 1760. Water power and then steam took work that had been done by human and animal muscle and gave it to machines. Textile mills, ironworks, railways.
The change was not only that things got faster. It was that output stopped being limited by how many people you could put in a building. Production could scale in a way it never had.
What made it possible was a new source of power that could be put wherever it was needed.
The second: scale and standardisation
From around 1870. Electricity, the assembly line, and the underrated part: interchangeable parts.
Electricity meant power could be distributed to each machine rather than transmitted by belts from one central engine, so factories could be laid out around the work instead of around the driveshaft. The assembly line meant work moved to the worker.
But the foundation was standardisation. Interchangeable parts are what make an assembly line possible at all: if every component has to be hand-fitted, there is no line. Mass production is a consequence of agreeing on specifications, not of conveyor belts.
That is the first appearance of a pattern worth holding on to: the enabling change was agreeing on a shared description of things, and the visible change came afterwards.
The third: each machine gets a controller
From around 1970. Electronics, computers, programmable logic controllers, robotics, and the first industrial IT.
Individual processes became automated. A control loop could hold a temperature without a person watching a gauge. A robot could weld the same seam ten thousand times. Historians began recording what happened, so operations could be reviewed rather than only remembered.
This is the revolution most industrial facilities are still living in, and it works well. But it has a specific shape worth naming: it automated machines one at a time. Each controller knows its own loop extremely well and knows nothing whatsoever about the machine next to it, the process they jointly serve, or the business objective either of them affects.
That limitation is exactly what the fourth revolution is about.
The fourth: the operation becomes something you can reason about
Now. The term comes from Industrie 4.0, a project in the German government's high-tech strategy, introduced publicly at the Hannover Fair in 2011. It has since been applied to almost everything, which has not helped anyone understand it.
Stripped of the marketing, the claim is specific: an operation that can be reasoned about as a whole system can sense what is happening, predict what is about to happen, and adjust itself, rather than depending on a person to notice, interpret and intervene.
The difference from the third revolution is not more automation. It is connection and context:
| Third revolution | Fourth revolution | |
|---|---|---|
| Scope | One machine, one loop | The operation as a system |
| Knows about | Its own setpoint | What it feeds, what feeds it, what depends on it |
| Reacts to | Its own measurement | Conditions elsewhere, and what they imply |
| Decides | Within a fixed rule | From context, within stated limits |
| Needs | A controller | A model of the operation |
What it actually requires
This is where most Industry 4.0 programmes go wrong, so it is worth being blunt.
The requirement is not sensors: they are cheap and you already have many. It is not connectivity, which is solved. It is not algorithms, which are open, well understood and mostly free.
The requirement is a machine-readable description of your operation: what exists, how it connects, and what it is a kind of. Without that, every clever technology above it is reasoning about numbers with no idea what they refer to. Kept in step with live data, that description has a familiar name: a digital twin.
Follow the historical pattern and this is unsurprising. Each revolution needed a foundation that looked boring next to the technology it enabled:
- The first needed power that could be placed where it was needed
- The second needed agreed specifications so parts were interchangeable
- The third needed computation cheap enough to put on every machine
- The fourth needs context, a shared, machine-readable model of the operation
Interchangeable parts were not the exciting part of the second revolution, and they were the part that made it work. The model plays the same role now.
Follow the trend line one step further and the operator stops being a person in a control room at all. Offshore, normally unmanned installations already run with no permanent crew, and a facility with nobody on board is operated through its digital twin, because the model is the only place the operation is visible. The foundation this page argues for is also that one. Where this is heading →
The operational data problem →
What this means practically
Three things follow, and they set the shape of a sensible programme.
Analytics, dashboards, predictive maintenance and AI agents all sit on top of it. Built afterwards, they each hand-assemble their own context and throw it away. Building your model →
The dominant cost is the time of people who know the operation, describing it. No vendor can supply that, because it is knowledge about your plant. Who does what →
Nobody becomes an Industry 4.0 operation in a quarter. You answer one question, model the slice it needs, and widen. Where to start → · Grown, not built →
"Industry 4.0" is a label, and a heavily marketed one. If it is unhelpful in your organisation, drop it: nothing on this site depends on the phrase. The capability underneath is real and specific: an operation with a written-down model of itself can be reasoned about by software. That is worth pursuing whatever it is called.
- The operational data problem: why the model is the missing piece
- What is a digital twin?: the name for the model the fourth revolution needs
- What is DataHub?: the plain-language explanation
- What is data liberation?: the first practical step
- AI agents: what becomes possible once the model exists
- Board briefing: this argument, condensed for governance