Have you ever thought about the difference between a tool and a machine? A recent exploration through the stack of books on my desk puzzled me with this thought.

I’ve been trying to understand how to frame the current AI Revolution. Not whether models are truly intelligent or whether AGI is around the corner. Those debates tend to miss the more useful question, which is structural: what kind of transformation are we actually in the middle of?

Looking for a better mental model, I ended up revisiting a book published in 1905 called The Industrial Revolution in the Eighteenth Century by Paul Mantoux.

Mantoux spends time examining a deceptively simple question: what separates a tool from a machine?

In his framing, a tool is something held in the worker’s hand. It amplifies human effort. A hammer, a chisel, a distaff. These extend what a skilled person can already do.

A machine, however, changes the relationship between the worker and the task. As Mantoux writes:

“Instead of being a tool in the workman’s hand, it is itself an artificial hand… a mechanism which, worked by any motive power, executes the elaborate movements of a technical operation which previously required one or several men.”

With a machine, the worker no longer performs the work directly. The worker supervises the mechanism performing the work.

Once Mantoux draws this distinction, he shows that machines rarely exist in isolation. They organize themselves into a larger structure.

Tools extend human skill. Machines replicate it. Factories coordinate machines. Industries coordinate factories.

The Industrial Revolution followed this progression. Individual machines appeared first. Those machines were then organized together inside factories. Over time, factories specialized and connected to form entire industries.

Mantoux obviously wasn’t writing about artificial intelligence, but the structure he described maps surprisingly well onto the transformation happening today.

The AI Revolution appears to be unfolding along a similar pattern.

Software has long functioned as our tool. It extends human cognition the way a hammer extends human labor. Software helps people do things faster and more efficiently, but the person still performs the work.

Agents begin to resemble machines.

An agent does not simply assist a human; it replicates a cognitive task. Through looping large language models, memory systems, and structured prompts, it executes work that previously required a person to perform step by step. In Mantoux’s terms, it functions as something close to an artificial hand for knowledge work.

At the moment, we are still firmly in the machine phase of this new revolution.

Agents today are inconsistent. They forget context, misinterpret instructions, and often require guardrails to stay on task. That shouldn’t surprise anyone who has looked closely at the early history of industrial machinery. The first generation of machines was also unreliable and required constant supervision.

Humans remained involved at the most critical points of those systems. Operators started the machines, corrected problems, and ensured they continued functioning. The same dynamic will exist with AI systems. Human judgment does not disappear; it simply moves to higher leverage points.

But what changed the industrial economy was not the existence of machines alone, but the ability for a single worker to supervise several machines simultaneously. Once that became possible, productivity expanded dramatically. A spinner who once operated a single wheel could eventually oversee multiple spinning frames.

A similar dynamic is beginning to appear in the AI Revolution.

One human supervising many agents. The Industrial Revolution scaled labor. The AI Revolution scales cognition.

But the larger transformation arrives with the next stage of the pattern: factories.

Mantoux argued that a factory should not be thought of primarily as a building. A factory is a coordinated system of machines powered by a shared source of energy. By bringing machines together into a single system, factories created scale, reliability, predictable output, and a structured division of labor.

The same architecture will emerge in AI.

Instead of factories made of machines, we will see cognitive factories: coordinated systems of agents executing defined cognitive functions. Each agent specializes in a particular task, and the system orchestrates those tasks together to produce consistent outcomes.

In the Industrial Revolution, the division of labor became encoded in machines.

In the AI Revolution, we are beginning to encode a division of cognition into agents.

From there, the pattern continues outward. Factories eventually coordinate with other factories, and industries emerge. The same dynamic will shape cognitive industries, where different cognitive factories specialize in particular forms of knowledge work. Some will optimize for speed. Others for cost. Others for accuracy or depth of reasoning.

Coordination between these systems will ultimately become one of the defining characteristics of the AI Revolution.

For now, however, the constraint is simpler.

Factories only work when their machines are reliable.

The same principle applies here. Cognitive factories require agents that can perform defined tasks consistently and repeatedly. Without that reliability, coordination becomes fragile, and large systems cannot form.

That is the real work happening today.

At Pinata, we’re focused on building agents that can perform defined cognitive tasks reliably and repeatably. Individual agents matter, but the real transformation will come from cognitive factories—systems where many agents coordinate together to perform complex knowledge work. Reliable agents are the foundation that makes those systems possible.

Over time, the most successful companies will not simply use agents; they will operate cognitive factories of their own.

If you're interested in building dependable agents that can evolve into cognitive factories inside your business—and generate real returns from AI—I’d love to talk.

Kyle Tut

Cofounder and CEO