Before the robot, fix the loom

According to a recent study, 64% of manufacturers that already have robots saying they are missing out on further opportunities but only 13% have an actual strategy for it. Most factories do not fail at automation because the technology is immature. They fail because they treat automation as a technical upgrade, not as a way of working. And increasingly, it is advanced human-machine collaboration that separates the few programs that scale from the many that stall. It is the deliberate design of how humans and machines sense problems, make decisions, learn, and improve together, day after day, line after line.

This is not a binary choice between people and robots but rather about deciding what the machine should do, what the human must do, and how the system improves when either one struggles. When that collaboration is designed explicitly, automation compounds. When it is left implicit, companies end up with isolated pilots, heroic operators, and impressive technology that never moves the P&L, despite years of investment and management attention. This condition is called pilot purgatory, a state where companies get stuck testing new technologies in small-scale pilots but fail to transition them into full-scale production.

Meanwhile, the global numbers keep moving. Robot density has doubled since 2017. Asia now installs more than 70% of all new industrial robots. China alone has climbed from fifth to third in robot density in just four years.

Wrong robots?

Contrary to popular belief, European manufacturers are not falling behind because they bought the wrong robots. In fact, Europe is among the most automated manufacturing regions in the world. The gap widens because many companies never built the operating system around those robots: the standards, roles, data flows, and routines that allow humans and machines to improve together and scale beyond the first line.

Whether it is called Lean, a production system, or operational excellence is secondary. What matters is having a stable, explicit system that makes automation repeatable. Without that foundation, automation remains stuck in pilots – quite impressive in isolation, but practically invisible on the P&L.

What Sakichi understood (and we tend to forget)

The idea is not new. Sakichi Toyoda’s loom was an early example of advanced human-machine collaboration, long before the term even existed. The machine stopped itself when quality faltered. The human diagnosed the cause. The system learned. Productivity, quality, and scalability followed.

A century later, the technology has changed – and quite significantly – but the principle remains the same: automation only scales when humans and machines are designed as a unified system.

The same patterns can be seen on the factory floor:

• Companies automate before they standardise. They place robots on top of processes with high variation, unclear standards, and multiple ways of running the same operation across shifts. The robot does not fix variation; it industrialises it, resulting in more waste.
• Companies buy automation as a product, not as an operating model. They focus on the cell, not the system. There is no data model, no operator routine, no playbook for the next line. The pilot works, but nothing else does.
• Companies design automation for operators, not with them. Sakichi's loom required a skilled human to diagnose what stopped it. Modern automation often treats the operator as a button-pusher. Adoption stalls, workarounds multiply, and the most experienced people are the first to leave.
• Companies measure technology readiness, not organisational readiness. They study the robot's payload and reach, then ignore whether the workforce, the maintenance team, the planners, and the leaders are ready to absorb what the robot brings.
• Companies confuse a pilot with a programme. A single successful cell is celebrated as a transformation. But years later, it remains exactly that: just one cell.

Every one of these failures would have been obvious to a 1924 weaving foreman. They are not technology problems. They are production system problems wearing technology clothes.

Do not automate chaos

Whether you call it Lean, a production system, or something else, it is the foundation that makes automation scalable. It is the discipline that makes human-machine collaboration explicit, repeatable, and effective. Advanced human-machine collaboration is not a philosophy layered on top of automation. It is the operating logic that makes automation worth scaling in the first place.

People are tempted to treat Lean as old-school – useful when the tools were paper and pencil, but much less relevant now that the tools are vision systems and AI. However, we argue that the opposite is true. The more advanced the automation, the higher the cost of automating an unstable process. The more sophisticated the technology, the bigger the gap between a pilot and a scaled programme.

Automation fails when there is no operating system to absorb it, not due to technological insufficiency. Clear standards, trusted data, and explicit human roles are what turn automation from isolated cells into scalable capability. Without that foundation, factories do not become future‑ready; they become showcases of individual automation potential, not general performance.

But discipline alone will not win the race. Incremental improvement cannot match competitors who are redesigning their factories around automation. The companies pulling ahead combine operational rigour with bold structural change – improving what exists while fundamentally rethinking how production is organised.

Across the automation programmes that actually scale, five principles show up again and again:

• Start with the business, not the technology. Automation is not the goal. Cost competitiveness, resilience, quality, and labour independence are. If you cannot draw a straight line from a robot to a number on the P&L, do not buy the robot yet.
• Standardise before you automate. Use Gemba, data, and co-creation to fix the process first. Sakichi did not put Jidoka on top of a chaotic loom; he redesigned the loom so the principle could live inside it. Do the same with yours.
• Design for human-machine collaboration from day one. Remember that operators are not a constraint on automation but the very reason it still works years later. Build their capability into the design, not after the fact.
• Build for the second line, not the first. Decide before the pilot how the solution will replicate across lines, shifts, and sites. Common data models, common standards, common operating routines. If you cannot see the path to line two, do not build line one.
• Sequence the waves. Wave one builds the foundation (stable processes, basic automation, reliable data). Wave two connects and coordinates. Wave three brings in AI and adaptive autonomy. Each wave earns the right to the next. Skipping waves is how you end up in pilot purgatory.

The thread is still breaking

A century after the Type G loom, the lesson has not changed. The point of Jidoka was never to remove the human but to make it matter more – to turn operators into problem-solvers, to make abnormality visible, to keep the system honest. The factories pulling ahead today are the ones that understand this line of thinking. They do not chase automation. Instead, they build the Lean foundation that makes automation worth chasing.

The 80% who are stuck are not stuck because they cannot make sense of the technology. They are stuck because they tried to automate before they were ready to be automated.

The good news is that the path forward is not new. It was written in a weaving shed in 1924, and it works just as well in a packaging line in 2026. You just have to be willing to stop the loom.