AI and the Age of Steam
AI and the Age of Steam
By Aaron Rose · Tech Reader Magazine · June 25, 2026
The Newcomen Engine
The Newcomen engine already existed before Watt touched it. A clumsy, atmospheric machine that worked by condensing steam inside a cylinder to create a vacuum — then letting air pressure push a piston down — it had been pumping water out of coal mines since around 1712. It worked, after a fashion, consuming enormous quantities of fuel to produce modest amounts of useful power.
Watt added a separate condenser so the cylinder never had to be cooled and reheated with each stroke, and the efficiency of the steam engine roughly quadrupled. What happened next is a lesson that neither the optimists nor the pessimists of any era have ever quite managed to learn: nobody knew what they had.
Not Watt. Not Boulton. Not the mine operators or textile merchants who would eventually transform Britain beyond recognition. They knew the machine was useful. They knew it could pump water and, later, drive a shaft. What they could not see — what no one could see — was that they were not holding a better pump. They were holding a new relationship between humanity and physical work.
They were holding a new relationship between humanity and physical work.
The Thing Nobody Saw Coming
For the first several decades after Watt's improvements, steam engines were used for exactly one thing: pumping water out of mines. This was the obvious application. Mines flooded. Steam could un-flood them. The fit was so clean that most people in the industry thought of steam as a mining technology, the way we might think of a backhoe as a construction technology. Useful. Specific. Limited.
The first factories to adopt steam power made a mistake so obvious in retrospect that it reads almost like comedy. They built their factories exactly as they had been built around water wheels. One central power source. A massive main shaft running through the building. Everything hanging off that shaft by belts and gears, arranged in a long row, all drawing from the same turning axis. They had swapped one power source for another and otherwise changed nothing at all.
It worked. Output increased. Costs fell. And for a generation, the factory owners thought that was the story: steam made the old way faster and cheaper. What they had not yet understood was that steam made the old way unnecessary.
It took roughly thirty years — some historians say longer — before factory designers began to realize that a steam engine did not have to be in the center of anything. It did not need to feed a single shaft. You could put a smaller engine wherever you needed power. You could arrange your factory around the work rather than around the machine. You could redesign everything. When that realization arrived, productivity did not increase. It exploded. The factory as a concept was rebuilt from scratch, and it bore almost no resemblance to what had come before.
What the World of 1800 Couldn't See
The predictions of the early industrial era read, from this distance, as a study in underestimation on both sides. The optimists saw cheaper goods and faster production. The pessimists saw dispossession and ruin. Both were right, and neither was close to capturing the full picture.
What the optimists missed was the challenge of the transition — the decades of displacement, the collapse of entire skilled trades that had taken generations to build, and a real human cost. The optimists were not wrong about the destination. They were wrong about the journey.
What the pessimists missed was everything that did not yet have a name. They could see the crafts that steam was destroying. They could not see the industries that did not yet exist — the locomotive engineers, the railway workers, the telegraph operators, the steel workers, the machine tool makers, the urban planners of the new industrial city. These were not variations on existing work. They were genuinely new categories of human activity that the world of 1800 had no language for, because the world of 1800 had no need for them.
Britain went from roughly ten million people in 1750 to twenty-one million by 1850 — and fed them all, employed most of them, and raised average living standards, however unequally distributed. Steam did not cause that. But steam made it possible in ways that its inventors never imagined and its critics never anticipated.
Steam created new categories of human activity that the world of 1800 had no language for.
Where the Age of Steam Resembles the Age of AI
The agentic AI moment of 2025 and 2026 resembles the early steam era in at least three ways that seem worth taking seriously.
The first way is the obvious: workflow automation. Just as the first factories used steam to power the same layout they had always used, the first wave of AI agent deployment is largely being used to automate the workflows that already exist. Write the report faster. Answer the emails more quickly. Schedule the meeting without the back-and-forth. These are genuine improvements. They are also, probably, the least interesting thing that agents will end up doing. We are wiring the new thing into the old building and calling it transformation.
The second way is in the invisibility of the new. The trades that steam destroyed were visible — you could walk into a workshop and see a handloom weaver and understand what he did. The industries that steam created were invisible until they existed. Nobody commissioned a study in 1790 on the future employment prospects in railway construction because there were no railways.
The jobs that agentic AI will create, if history is a guide, will not be variations on what exists today. They will be things that do not yet have names, for problems that have not yet been identified, in industries that have not yet been founded.
The third way involves the change itself. The transition from handloom to power loom took roughly forty years in Britain, and it was, for many of the people who lived through it, genuinely difficult. The fact that the destination turned out to be better — by most measures, for most people, over the long arc — did not make the journey easier. History has a way of glossing over the fact that adapting to the age of steam was difficult for many people at the time.
The jobs that agentic AI create will not be variations on what exists today. They will be jobs that do not yet have names, for problems that have not yet been identified, in industries that have not yet been founded.
Where the Similarities Break Down
The rate of change in the age of steam was slow. This is the most important fact about it, and it may be where the analogy between steam and agentic AI begins to fracture.
The Watt engine was improved in 1769. The railways did not arrive until the 1820s and 1830s. The full redesign of factory architecture took decades. The social and economic disruptions of industrialization played out over generations.
Workers had decades to adapt, even if many of them adapted slowly or not at all. Institutions had generations to respond. Governments, unions, education systems, and social safety nets — however inadequate — had the luxury of catching up slowly to a technology that was itself moving slowly.
Agentic AI is not moving slowly. The trajectory from large language models to autonomous agents capable of executing complex multi-step tasks has compressed what might have been a decade of development into something closer to two or three years.
Whether that pace continues, accelerates, or plateaus is genuinely unknown. But if the parallel of AI with steam holds in the broad strokes — disruption, displacement, and eventual transformation — and if the pace of change is radically faster, then human institutions will have to adapt quickly along with AI.
The Question History Can't Answer
Here is what steam can tell us: general-purpose technologies are almost always underestimated in their eventual scope and overestimated in their short-term elegance. They change things that are visible before they create things that are not yet nameable. They are adopted wrong before they are adopted right.
The delay between steam's arrival and the factory redesign that unlocked its real potential is not an anomaly. It is, historically speaking, the normal way that civilization processes a genuinely new kind of power.
Here is what steam cannot tell us: how society can keep up with the rapid pace of change brought about by agentic AI. Certainly, time-frames are compressed today compared to the 1700s and 1800s. And rapid change is not restricted to the world of AI. It's everywhere now. Agentic AI will probably lead to new opportunities and markets in a matter of months, compared to the decades of development required with steam.
We will probably see new opportunities and markets develop in a matter of months with agentic AI, compared to the decades of development required with steam.
These are not rhetorical questions. They are open ones. The age of steam resembles agentic AI in ways that are instructive and clarifying. It does not resemble it so perfectly that we can read the future exactly from the past.
Coming Next: The Spreadsheet
Everyone predicted accountants would disappear. Instead, accounting exploded, finance expanded, and analysts multiplied. The spreadsheet is the closest historical analogy to agentic AI — and its story is more surprising than the prediction ever was.
Tech Reader Magazine
TechReaderMagazine.com