Human-centered automation starts with one question. When I evaluate an automation proposal, I want to know how it will make the operation more capable.
Will quality decisions become more dependable? Will experienced people have more time for troubleshooting and improvement? Will the factory handle variation better? Who will take responsibility when the system encounters something it does not understand?
These questions give human-centered automation substance. They also shape how I think about Industry 5.0: technology should amplify manufacturing expertise, with a deliberate division of responsibility between people and machines.
For leaders investing in AI and automation, I believe five demands deserve a place in the conversation.
1. Human-centered automation starts with the actual work
One customer described a fastening operation where different positions on an assembly required different programmed torque settings. The controller expected the operator to follow a particular sequence.
The physical work did not always follow that sequence. An operator might move ahead or return to a location, leaving the controller expecting a different fastening position.
Calling this human error would leave an important engineering question unanswered: why had the digital system lost track of the physical work?
We discussed using vision to identify the operator’s fastening position and communicate the corresponding program to the torque-control process. The deployment details and results are confidential, so I attach no performance claim to that example.
The principle is useful: leaders should demand that the design account for the real workflow. Where sequence is essential to quality or safety, the process needs appropriate controls. Where legitimate variation is possible, the system should account for it.
2. Define what the technology is authorized to decide
A vision system’s ability to observe a condition does not automatically give it authority to determine the production response.
An approved inspection application may recognize a known defect and trigger an agreed action. An unfamiliar condition may require qualified review. A recurring pattern may justify investigation without establishing its cause.
Before deployment, leaders should ask:
- What can the system observe?
- What may it interpret from that evidence?
- What action may it trigger?
- Who may change the rules?
Clear answers make automation easier to govern and give people a practical understanding of their responsibilities.
3. Explain what happens when the normal process breaks
A demonstration usually shows the expected sequence. Production will eventually test the exceptions.
What happens when the camera cannot capture usable evidence? When the controller is unavailable? When the component differs from the expected configuration? When an operator challenges the automated guidance?
The response might involve holding the operation, requesting review, using an approved fallback, or stopping the sequence. The appropriate choice depends on the application.
Leaders should demand an intentional exception path, with named owners and clear authority. Production pressure is a poor setting for inventing those arrangements.
Human review can be part of a well-designed automated process. Its purpose and limits should be established before the line depends on it.
4. Measure where the work moved
Automating a task changes the distribution of work.
Repetitive inspection may decrease while maintenance, secondary review, training, or technical support increases. An experienced operator may gain time for setup and troubleshooting. A quality engineer may spend less time sorting components and more time investigating recurring conditions.
Those changes can create substantial value, but they need to be measured.
I want to understand whether the project releases useful capacity, improves quality, or allows scarce expertise to contribute more. I also want the business case to include the new work the system creates.
Headcount alone cannot describe that outcome.
5. Human-centered automation preserves the factory’s ability to learn
Products, suppliers, equipment, and operating conditions change after installation.
The people closest to production often recognize when those changes challenge an assumption behind the system. Their observations need a route into investigation, review, and approved improvement.
Leaders should demand a process for capturing evidence, examining disagreements, and deciding when revalidation is necessary.
That is how automation remains useful as the factory changes.
My test for human-centered automation is practical: has the complete operation become more capable, and can the people responsible for it understand, support, and improve it?
For more on this approach, read Chapter 19, “Industry 5.0 Is Human Expertise Amplified,” in my book, Teach the Factory to See. Get it on Amazon.
